tcg/ppc/tcg-target.c.inc

/*
 * Tiny Code Generator for QEMU
 *
 * Copyright (c) 2008 Fabrice Bellard
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 * THE SOFTWARE.
 */

#include "elf.h"

/*
 * Standardize on the _CALL_FOO symbols used by GCC:
 * Apple XCode does not define _CALL_DARWIN.
 * Clang defines _CALL_ELF (64-bit) but not _CALL_SYSV or _CALL_AIX.
 */
#if TCG_TARGET_REG_BITS == 64
# ifdef _CALL_AIX
    /* ok */
# elif defined(_CALL_ELF) && _CALL_ELF == 1
#  define _CALL_AIX
# elif defined(_CALL_ELF) && _CALL_ELF == 2
    /* ok */
# else
#  error "Unknown ABI"
# endif
#else
# if defined(_CALL_SYSV) || defined(_CALL_DARWIN)
    /* ok */
# elif defined(__APPLE__)
#  define _CALL_DARWIN
# elif defined(__ELF__)
#  define _CALL_SYSV
# else
#  error "Unknown ABI"
# endif
#endif

#if TCG_TARGET_REG_BITS == 64
# define TCG_TARGET_CALL_ARG_I32   TCG_CALL_ARG_EXTEND
# define TCG_TARGET_CALL_RET_I128  TCG_CALL_RET_NORMAL
#else
# define TCG_TARGET_CALL_ARG_I32   TCG_CALL_ARG_NORMAL
# define TCG_TARGET_CALL_RET_I128  TCG_CALL_RET_BY_REF
#endif
#ifdef _CALL_SYSV
# define TCG_TARGET_CALL_ARG_I64   TCG_CALL_ARG_EVEN
# define TCG_TARGET_CALL_ARG_I128  TCG_CALL_ARG_BY_REF
#else
# define TCG_TARGET_CALL_ARG_I64   TCG_CALL_ARG_NORMAL
# define TCG_TARGET_CALL_ARG_I128  TCG_CALL_ARG_NORMAL
#endif

/* For some memory operations, we need a scratch that isn't R0.  For the AIX
   calling convention, we can re-use the TOC register since we'll be reloading
   it at every call.  Otherwise R12 will do nicely as neither a call-saved
   register nor a parameter register.  */
#ifdef _CALL_AIX
# define TCG_REG_TMP1   TCG_REG_R2
#else
# define TCG_REG_TMP1   TCG_REG_R12
#endif
#define TCG_REG_TMP2    TCG_REG_R11

#define TCG_VEC_TMP1    TCG_REG_V0
#define TCG_VEC_TMP2    TCG_REG_V1

#define TCG_REG_TB     TCG_REG_R31
#define USE_REG_TB     (TCG_TARGET_REG_BITS == 64 && !have_isa_3_00)

/* Shorthand for size of a pointer.  Avoid promotion to unsigned.  */
#define SZP  ((int)sizeof(void *))

/* Shorthand for size of a register.  */
#define SZR  (TCG_TARGET_REG_BITS / 8)

#define TCG_CT_CONST_S16  0x100
#define TCG_CT_CONST_U16  0x200
#define TCG_CT_CONST_S32  0x400
#define TCG_CT_CONST_U32  0x800
#define TCG_CT_CONST_ZERO 0x1000
#define TCG_CT_CONST_MONE 0x2000
#define TCG_CT_CONST_WSZ  0x4000
#define TCG_CT_CONST_CMP  0x8000

#define ALL_GENERAL_REGS  0xffffffffu
#define ALL_VECTOR_REGS   0xffffffff00000000ull

#ifndef R_PPC64_PCREL34
#define R_PPC64_PCREL34  132
#endif

#define have_isel  (cpuinfo & CPUINFO_ISEL)

#define TCG_GUEST_BASE_REG  TCG_REG_R30

#ifdef CONFIG_DEBUG_TCG
static const char tcg_target_reg_names[TCG_TARGET_NB_REGS][4] = {
    "r0",  "r1",  "r2",  "r3",  "r4",  "r5",  "r6",  "r7",
    "r8",  "r9",  "r10", "r11", "r12", "r13", "r14", "r15",
    "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
    "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31",
    "v0",  "v1",  "v2",  "v3",  "v4",  "v5",  "v6",  "v7",
    "v8",  "v9",  "v10", "v11", "v12", "v13", "v14", "v15",
    "v16", "v17", "v18", "v19", "v20", "v21", "v22", "v23",
    "v24", "v25", "v26", "v27", "v28", "v29", "v30", "v31",
};
#endif

static const int tcg_target_reg_alloc_order[] = {
    TCG_REG_R14,  /* call saved registers */
    TCG_REG_R15,
    TCG_REG_R16,
    TCG_REG_R17,
    TCG_REG_R18,
    TCG_REG_R19,
    TCG_REG_R20,
    TCG_REG_R21,
    TCG_REG_R22,
    TCG_REG_R23,
    TCG_REG_R24,
    TCG_REG_R25,
    TCG_REG_R26,
    TCG_REG_R27,
    TCG_REG_R28,
    TCG_REG_R29,
    TCG_REG_R30,
    TCG_REG_R31,
    TCG_REG_R12,  /* call clobbered, non-arguments */
    TCG_REG_R11,
    TCG_REG_R2,
    TCG_REG_R13,
    TCG_REG_R10,  /* call clobbered, arguments */
    TCG_REG_R9,
    TCG_REG_R8,
    TCG_REG_R7,
    TCG_REG_R6,
    TCG_REG_R5,
    TCG_REG_R4,
    TCG_REG_R3,

    /* V0 and V1 reserved as temporaries; V20 - V31 are call-saved */
    TCG_REG_V2,   /* call clobbered, vectors */
    TCG_REG_V3,
    TCG_REG_V4,
    TCG_REG_V5,
    TCG_REG_V6,
    TCG_REG_V7,
    TCG_REG_V8,
    TCG_REG_V9,
    TCG_REG_V10,
    TCG_REG_V11,
    TCG_REG_V12,
    TCG_REG_V13,
    TCG_REG_V14,
    TCG_REG_V15,
    TCG_REG_V16,
    TCG_REG_V17,
    TCG_REG_V18,
    TCG_REG_V19,
};

static const int tcg_target_call_iarg_regs[] = {
    TCG_REG_R3,
    TCG_REG_R4,
    TCG_REG_R5,
    TCG_REG_R6,
    TCG_REG_R7,
    TCG_REG_R8,
    TCG_REG_R9,
    TCG_REG_R10
};

static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
{
    tcg_debug_assert(kind == TCG_CALL_RET_NORMAL);
    tcg_debug_assert(slot >= 0 && slot <= 1);
    return TCG_REG_R3 + slot;
}

static const int tcg_target_callee_save_regs[] = {
#ifdef _CALL_DARWIN
    TCG_REG_R11,
#endif
    TCG_REG_R14,
    TCG_REG_R15,
    TCG_REG_R16,
    TCG_REG_R17,
    TCG_REG_R18,
    TCG_REG_R19,
    TCG_REG_R20,
    TCG_REG_R21,
    TCG_REG_R22,
    TCG_REG_R23,
    TCG_REG_R24,
    TCG_REG_R25,
    TCG_REG_R26,
    TCG_REG_R27, /* currently used for the global env */
    TCG_REG_R28,
    TCG_REG_R29,
    TCG_REG_R30,
    TCG_REG_R31
};

/* For PPC, we use TB+4 instead of TB as the base. */
static inline ptrdiff_t ppc_tbrel_diff(TCGContext *s, const void *target)
{
    return tcg_tbrel_diff(s, target) - 4;
}

static inline bool in_range_b(tcg_target_long target)
{
    return target == sextract64(target, 0, 26);
}

static uint32_t reloc_pc24_val(const tcg_insn_unit *pc,
                               const tcg_insn_unit *target)
{
    ptrdiff_t disp = tcg_ptr_byte_diff(target, pc);
    tcg_debug_assert(in_range_b(disp));
    return disp & 0x3fffffc;
}

static bool reloc_pc24(tcg_insn_unit *src_rw, const tcg_insn_unit *target)
{
    const tcg_insn_unit *src_rx = tcg_splitwx_to_rx(src_rw);
    ptrdiff_t disp = tcg_ptr_byte_diff(target, src_rx);

    if (in_range_b(disp)) {
        *src_rw = (*src_rw & ~0x3fffffc) | (disp & 0x3fffffc);
        return true;
    }
    return false;
}

static uint16_t reloc_pc14_val(const tcg_insn_unit *pc,
                               const tcg_insn_unit *target)
{
    ptrdiff_t disp = tcg_ptr_byte_diff(target, pc);
    tcg_debug_assert(disp == (int16_t) disp);
    return disp & 0xfffc;
}

static bool reloc_pc14(tcg_insn_unit *src_rw, const tcg_insn_unit *target)
{
    const tcg_insn_unit *src_rx = tcg_splitwx_to_rx(src_rw);
    ptrdiff_t disp = tcg_ptr_byte_diff(target, src_rx);

    if (disp == (int16_t) disp) {
        *src_rw = (*src_rw & ~0xfffc) | (disp & 0xfffc);
        return true;
    }
    return false;
}

static bool reloc_pc34(tcg_insn_unit *src_rw, const tcg_insn_unit *target)
{
    const tcg_insn_unit *src_rx = tcg_splitwx_to_rx(src_rw);
    ptrdiff_t disp = tcg_ptr_byte_diff(target, src_rx);

    if (disp == sextract64(disp, 0, 34)) {
        src_rw[0] = (src_rw[0] & ~0x3ffff) | ((disp >> 16) & 0x3ffff);
        src_rw[1] = (src_rw[1] & ~0xffff) | (disp & 0xffff);
        return true;
    }
    return false;
}

static bool mask_operand(uint32_t c, int *mb, int *me);
static bool mask64_operand(uint64_t c, int *mb, int *me);

/* test if a constant matches the constraint */
static bool tcg_target_const_match(int64_t sval, int ct,
                                   TCGType type, TCGCond cond, int vece)
{
    uint64_t uval = sval;
    int mb, me;

    if (ct & TCG_CT_CONST) {
        return 1;
    }

    if (type == TCG_TYPE_I32) {
        uval = (uint32_t)sval;
        sval = (int32_t)sval;
    }

    if (ct & TCG_CT_CONST_CMP) {
        switch (cond) {
        case TCG_COND_EQ:
        case TCG_COND_NE:
            ct |= TCG_CT_CONST_S16 | TCG_CT_CONST_U16;
            break;
        case TCG_COND_LT:
        case TCG_COND_GE:
        case TCG_COND_LE:
        case TCG_COND_GT:
            ct |= TCG_CT_CONST_S16;
            break;
        case TCG_COND_LTU:
        case TCG_COND_GEU:
        case TCG_COND_LEU:
        case TCG_COND_GTU:
            ct |= TCG_CT_CONST_U16;
            break;
        case TCG_COND_TSTEQ:
        case TCG_COND_TSTNE:
            if ((uval & ~0xffff) == 0 || (uval & ~0xffff0000ull) == 0) {
                return 1;
            }
            if (uval == (uint32_t)uval && mask_operand(uval, &mb, &me)) {
                return 1;
            }
            if (TCG_TARGET_REG_BITS == 64 &&
                mask64_operand(uval << clz64(uval), &mb, &me)) {
                return 1;
            }
            return 0;
        default:
            g_assert_not_reached();
        }
    }

    if ((ct & TCG_CT_CONST_S16) && sval == (int16_t)sval) {
        return 1;
    }
    if ((ct & TCG_CT_CONST_U16) && uval == (uint16_t)uval) {
        return 1;
    }
    if ((ct & TCG_CT_CONST_S32) && sval == (int32_t)sval) {
        return 1;
    }
    if ((ct & TCG_CT_CONST_U32) && uval == (uint32_t)uval) {
        return 1;
    }
    if ((ct & TCG_CT_CONST_ZERO) && sval == 0) {
        return 1;
    }
    if ((ct & TCG_CT_CONST_MONE) && sval == -1) {
        return 1;
    }
    if ((ct & TCG_CT_CONST_WSZ) && sval == (type == TCG_TYPE_I32 ? 32 : 64)) {
        return 1;
    }
    return 0;
}

#define OPCD(opc) ((opc)<<26)
#define XO19(opc) (OPCD(19)|((opc)<<1))
#define MD30(opc) (OPCD(30)|((opc)<<2))
#define MDS30(opc) (OPCD(30)|((opc)<<1))
#define XO31(opc) (OPCD(31)|((opc)<<1))
#define XO58(opc) (OPCD(58)|(opc))
#define XO62(opc) (OPCD(62)|(opc))
#define VX4(opc)  (OPCD(4)|(opc))

#define B      OPCD( 18)
#define BC     OPCD( 16)

#define LBZ    OPCD( 34)
#define LHZ    OPCD( 40)
#define LHA    OPCD( 42)
#define LWZ    OPCD( 32)
#define LWZUX  XO31( 55)
#define LD     XO58(  0)
#define LDX    XO31( 21)
#define LDU    XO58(  1)
#define LDUX   XO31( 53)
#define LWA    XO58(  2)
#define LWAX   XO31(341)
#define LQ     OPCD( 56)

#define STB    OPCD( 38)
#define STH    OPCD( 44)
#define STW    OPCD( 36)
#define STD    XO62(  0)
#define STDU   XO62(  1)
#define STDX   XO31(149)
#define STQ    XO62(  2)

#define PLWA   OPCD( 41)
#define PLD    OPCD( 57)
#define PLXSD  OPCD( 42)
#define PLXV   OPCD(25 * 2 + 1)  /* force tx=1 */

#define PSTD   OPCD( 61)
#define PSTXSD OPCD( 46)
#define PSTXV  OPCD(27 * 2 + 1)  /* force sx=1 */

#define ADDIC  OPCD( 12)
#define ADDI   OPCD( 14)
#define ADDIS  OPCD( 15)
#define ORI    OPCD( 24)
#define ORIS   OPCD( 25)
#define XORI   OPCD( 26)
#define XORIS  OPCD( 27)
#define ANDI   OPCD( 28)
#define ANDIS  OPCD( 29)
#define MULLI  OPCD(  7)
#define CMPLI  OPCD( 10)
#define CMPI   OPCD( 11)
#define SUBFIC OPCD( 8)

#define LWZU   OPCD( 33)
#define STWU   OPCD( 37)

#define RLWIMI OPCD( 20)
#define RLWINM OPCD( 21)
#define RLWNM  OPCD( 23)

#define RLDICL MD30(  0)
#define RLDICR MD30(  1)
#define RLDIMI MD30(  3)
#define RLDCL  MDS30( 8)

#define BCLR   XO19( 16)
#define BCCTR  XO19(528)
#define CRAND  XO19(257)
#define CRANDC XO19(129)
#define CRNAND XO19(225)
#define CROR   XO19(449)
#define CRNOR  XO19( 33)
#define ADDPCIS XO19( 2)

#define EXTSB  XO31(954)
#define EXTSH  XO31(922)
#define EXTSW  XO31(986)
#define ADD    XO31(266)
#define ADDE   XO31(138)
#define ADDME  XO31(234)
#define ADDZE  XO31(202)
#define ADDC   XO31( 10)
#define AND    XO31( 28)
#define SUBF   XO31( 40)
#define SUBFC  XO31(  8)
#define SUBFE  XO31(136)
#define SUBFME XO31(232)
#define SUBFZE XO31(200)
#define OR     XO31(444)
#define XOR    XO31(316)
#define MULLW  XO31(235)
#define MULHW  XO31( 75)
#define MULHWU XO31( 11)
#define DIVW   XO31(491)
#define DIVWU  XO31(459)
#define MODSW  XO31(779)
#define MODUW  XO31(267)
#define CMP    XO31(  0)
#define CMPL   XO31( 32)
#define LHBRX  XO31(790)
#define LWBRX  XO31(534)
#define LDBRX  XO31(532)
#define STHBRX XO31(918)
#define STWBRX XO31(662)
#define STDBRX XO31(660)
#define MFSPR  XO31(339)
#define MTSPR  XO31(467)
#define SRAWI  XO31(824)
#define NEG    XO31(104)
#define MFCR   XO31( 19)
#define MFOCRF (MFCR | (1u << 20))
#define NOR    XO31(124)
#define CNTLZW XO31( 26)
#define CNTLZD XO31( 58)
#define CNTTZW XO31(538)
#define CNTTZD XO31(570)
#define CNTPOPW XO31(378)
#define CNTPOPD XO31(506)
#define ANDC   XO31( 60)
#define ORC    XO31(412)
#define EQV    XO31(284)
#define NAND   XO31(476)
#define ISEL   XO31( 15)

#define MULLD  XO31(233)
#define MULHD  XO31( 73)
#define MULHDU XO31(  9)
#define DIVD   XO31(489)
#define DIVDU  XO31(457)
#define MODSD  XO31(777)
#define MODUD  XO31(265)

#define LBZX   XO31( 87)
#define LHZX   XO31(279)
#define LHAX   XO31(343)
#define LWZX   XO31( 23)
#define STBX   XO31(215)
#define STHX   XO31(407)
#define STWX   XO31(151)

#define EIEIO  XO31(854)
#define HWSYNC XO31(598)
#define LWSYNC (HWSYNC | (1u << 21))

#define SPR(a, b) ((((a)<<5)|(b))<<11)
#define LR     SPR(8, 0)
#define CTR    SPR(9, 0)

#define SLW    XO31( 24)
#define SRW    XO31(536)
#define SRAW   XO31(792)

#define SLD    XO31( 27)
#define SRD    XO31(539)
#define SRAD   XO31(794)
#define SRADI  XO31(413<<1)

#define BRH    XO31(219)
#define BRW    XO31(155)
#define BRD    XO31(187)

#define TW     XO31( 4)
#define TRAP   (TW | TO(31))

#define SETBC    XO31(384)  /* v3.10 */
#define SETBCR   XO31(416)  /* v3.10 */
#define SETNBC   XO31(448)  /* v3.10 */
#define SETNBCR  XO31(480)  /* v3.10 */

#define NOP    ORI  /* ori 0,0,0 */

#define LVX        XO31(103)
#define LVEBX      XO31(7)
#define LVEHX      XO31(39)
#define LVEWX      XO31(71)
#define LXSDX      (XO31(588) | 1)  /* v2.06, force tx=1 */
#define LXVDSX     (XO31(332) | 1)  /* v2.06, force tx=1 */
#define LXSIWZX    (XO31(12) | 1)   /* v2.07, force tx=1 */
#define LXV        (OPCD(61) | 8 | 1)  /* v3.00, force tx=1 */
#define LXSD       (OPCD(57) | 2)   /* v3.00 */
#define LXVWSX     (XO31(364) | 1)  /* v3.00, force tx=1 */

#define STVX       XO31(231)
#define STVEWX     XO31(199)
#define STXSDX     (XO31(716) | 1)  /* v2.06, force sx=1 */
#define STXSIWX    (XO31(140) | 1)  /* v2.07, force sx=1 */
#define STXV       (OPCD(61) | 8 | 5) /* v3.00, force sx=1 */
#define STXSD      (OPCD(61) | 2)   /* v3.00 */

#define VADDSBS    VX4(768)
#define VADDUBS    VX4(512)
#define VADDUBM    VX4(0)
#define VADDSHS    VX4(832)
#define VADDUHS    VX4(576)
#define VADDUHM    VX4(64)
#define VADDSWS    VX4(896)
#define VADDUWS    VX4(640)
#define VADDUWM    VX4(128)
#define VADDUDM    VX4(192)       /* v2.07 */

#define VSUBSBS    VX4(1792)
#define VSUBUBS    VX4(1536)
#define VSUBUBM    VX4(1024)
#define VSUBSHS    VX4(1856)
#define VSUBUHS    VX4(1600)
#define VSUBUHM    VX4(1088)
#define VSUBSWS    VX4(1920)
#define VSUBUWS    VX4(1664)
#define VSUBUWM    VX4(1152)
#define VSUBUDM    VX4(1216)      /* v2.07 */

#define VNEGW      (VX4(1538) | (6 << 16))  /* v3.00 */
#define VNEGD      (VX4(1538) | (7 << 16))  /* v3.00 */

#define VMAXSB     VX4(258)
#define VMAXSH     VX4(322)
#define VMAXSW     VX4(386)
#define VMAXSD     VX4(450)       /* v2.07 */
#define VMAXUB     VX4(2)
#define VMAXUH     VX4(66)
#define VMAXUW     VX4(130)
#define VMAXUD     VX4(194)       /* v2.07 */
#define VMINSB     VX4(770)
#define VMINSH     VX4(834)
#define VMINSW     VX4(898)
#define VMINSD     VX4(962)       /* v2.07 */
#define VMINUB     VX4(514)
#define VMINUH     VX4(578)
#define VMINUW     VX4(642)
#define VMINUD     VX4(706)       /* v2.07 */

#define VCMPEQUB   VX4(6)
#define VCMPEQUH   VX4(70)
#define VCMPEQUW   VX4(134)
#define VCMPEQUD   VX4(199)       /* v2.07 */
#define VCMPGTSB   VX4(774)
#define VCMPGTSH   VX4(838)
#define VCMPGTSW   VX4(902)
#define VCMPGTSD   VX4(967)       /* v2.07 */
#define VCMPGTUB   VX4(518)
#define VCMPGTUH   VX4(582)
#define VCMPGTUW   VX4(646)
#define VCMPGTUD   VX4(711)       /* v2.07 */
#define VCMPNEB    VX4(7)         /* v3.00 */
#define VCMPNEH    VX4(71)        /* v3.00 */
#define VCMPNEW    VX4(135)       /* v3.00 */

#define VSLB       VX4(260)
#define VSLH       VX4(324)
#define VSLW       VX4(388)
#define VSLD       VX4(1476)      /* v2.07 */
#define VSRB       VX4(516)
#define VSRH       VX4(580)
#define VSRW       VX4(644)
#define VSRD       VX4(1732)      /* v2.07 */
#define VSRAB      VX4(772)
#define VSRAH      VX4(836)
#define VSRAW      VX4(900)
#define VSRAD      VX4(964)       /* v2.07 */
#define VRLB       VX4(4)
#define VRLH       VX4(68)
#define VRLW       VX4(132)
#define VRLD       VX4(196)       /* v2.07 */

#define VMULEUB    VX4(520)
#define VMULEUH    VX4(584)
#define VMULEUW    VX4(648)       /* v2.07 */
#define VMULOUB    VX4(8)
#define VMULOUH    VX4(72)
#define VMULOUW    VX4(136)       /* v2.07 */
#define VMULUWM    VX4(137)       /* v2.07 */
#define VMULLD     VX4(457)       /* v3.10 */
#define VMSUMUHM   VX4(38)

#define VMRGHB     VX4(12)
#define VMRGHH     VX4(76)
#define VMRGHW     VX4(140)
#define VMRGLB     VX4(268)
#define VMRGLH     VX4(332)
#define VMRGLW     VX4(396)

#define VPKUHUM    VX4(14)
#define VPKUWUM    VX4(78)

#define VAND       VX4(1028)
#define VANDC      VX4(1092)
#define VNOR       VX4(1284)
#define VOR        VX4(1156)
#define VXOR       VX4(1220)
#define VEQV       VX4(1668)      /* v2.07 */
#define VNAND      VX4(1412)      /* v2.07 */
#define VORC       VX4(1348)      /* v2.07 */

#define VSPLTB     VX4(524)
#define VSPLTH     VX4(588)
#define VSPLTW     VX4(652)
#define VSPLTISB   VX4(780)
#define VSPLTISH   VX4(844)
#define VSPLTISW   VX4(908)

#define VSLDOI     VX4(44)

#define XXPERMDI   (OPCD(60) | (10 << 3) | 7)  /* v2.06, force ax=bx=tx=1 */
#define XXSEL      (OPCD(60) | (3 << 4) | 0xf) /* v2.06, force ax=bx=cx=tx=1 */
#define XXSPLTIB   (OPCD(60) | (360 << 1) | 1) /* v3.00, force tx=1 */

#define MFVSRD     (XO31(51) | 1)   /* v2.07, force sx=1 */
#define MFVSRWZ    (XO31(115) | 1)  /* v2.07, force sx=1 */
#define MTVSRD     (XO31(179) | 1)  /* v2.07, force tx=1 */
#define MTVSRWZ    (XO31(243) | 1)  /* v2.07, force tx=1 */
#define MTVSRDD    (XO31(435) | 1)  /* v3.00, force tx=1 */
#define MTVSRWS    (XO31(403) | 1)  /* v3.00, force tx=1 */

#define RT(r) ((r)<<21)
#define RS(r) ((r)<<21)
#define RA(r) ((r)<<16)
#define RB(r) ((r)<<11)
#define TO(t) ((t)<<21)
#define SH(s) ((s)<<11)
#define MB(b) ((b)<<6)
#define ME(e) ((e)<<1)
#define BO(o) ((o)<<21)
#define MB64(b) ((b)<<5)
#define FXM(b) (1 << (19 - (b)))

#define VRT(r)  (((r) & 31) << 21)
#define VRA(r)  (((r) & 31) << 16)
#define VRB(r)  (((r) & 31) << 11)
#define VRC(r)  (((r) & 31) <<  6)

#define LK    1

#define TAB(t, a, b) (RT(t) | RA(a) | RB(b))
#define SAB(s, a, b) (RS(s) | RA(a) | RB(b))
#define TAI(s, a, i) (RT(s) | RA(a) | ((i) & 0xffff))
#define SAI(s, a, i) (RS(s) | RA(a) | ((i) & 0xffff))

#define BF(n)    ((n)<<23)
#define BI(n, c) (((c)+((n)*4))<<16)
#define BT(n, c) (((c)+((n)*4))<<21)
#define BA(n, c) (((c)+((n)*4))<<16)
#define BB(n, c) (((c)+((n)*4))<<11)
#define BC_(n, c) (((c)+((n)*4))<<6)

#define BO_COND_TRUE  BO(12)
#define BO_COND_FALSE BO( 4)
#define BO_ALWAYS     BO(20)

enum {
    CR_LT,
    CR_GT,
    CR_EQ,
    CR_SO
};

static const uint32_t tcg_to_bc[16] = {
    [TCG_COND_EQ]  = BC | BI(0, CR_EQ) | BO_COND_TRUE,
    [TCG_COND_NE]  = BC | BI(0, CR_EQ) | BO_COND_FALSE,
    [TCG_COND_TSTEQ]  = BC | BI(0, CR_EQ) | BO_COND_TRUE,
    [TCG_COND_TSTNE]  = BC | BI(0, CR_EQ) | BO_COND_FALSE,
    [TCG_COND_LT]  = BC | BI(0, CR_LT) | BO_COND_TRUE,
    [TCG_COND_GE]  = BC | BI(0, CR_LT) | BO_COND_FALSE,
    [TCG_COND_LE]  = BC | BI(0, CR_GT) | BO_COND_FALSE,
    [TCG_COND_GT]  = BC | BI(0, CR_GT) | BO_COND_TRUE,
    [TCG_COND_LTU] = BC | BI(0, CR_LT) | BO_COND_TRUE,
    [TCG_COND_GEU] = BC | BI(0, CR_LT) | BO_COND_FALSE,
    [TCG_COND_LEU] = BC | BI(0, CR_GT) | BO_COND_FALSE,
    [TCG_COND_GTU] = BC | BI(0, CR_GT) | BO_COND_TRUE,
};

/* The low bit here is set if the RA and RB fields must be inverted.  */
static const uint32_t tcg_to_isel[16] = {
    [TCG_COND_EQ]  = ISEL | BC_(0, CR_EQ),
    [TCG_COND_NE]  = ISEL | BC_(0, CR_EQ) | 1,
    [TCG_COND_TSTEQ] = ISEL | BC_(0, CR_EQ),
    [TCG_COND_TSTNE] = ISEL | BC_(0, CR_EQ) | 1,
    [TCG_COND_LT]  = ISEL | BC_(0, CR_LT),
    [TCG_COND_GE]  = ISEL | BC_(0, CR_LT) | 1,
    [TCG_COND_LE]  = ISEL | BC_(0, CR_GT) | 1,
    [TCG_COND_GT]  = ISEL | BC_(0, CR_GT),
    [TCG_COND_LTU] = ISEL | BC_(0, CR_LT),
    [TCG_COND_GEU] = ISEL | BC_(0, CR_LT) | 1,
    [TCG_COND_LEU] = ISEL | BC_(0, CR_GT) | 1,
    [TCG_COND_GTU] = ISEL | BC_(0, CR_GT),
};

static bool patch_reloc(tcg_insn_unit *code_ptr, int type,
                        intptr_t value, intptr_t addend)
{
    const tcg_insn_unit *target;
    int16_t lo;
    int32_t hi;

    value += addend;
    target = (const tcg_insn_unit *)value;

    switch (type) {
    case R_PPC_REL14:
        return reloc_pc14(code_ptr, target);
    case R_PPC_REL24:
        return reloc_pc24(code_ptr, target);
    case R_PPC64_PCREL34:
        return reloc_pc34(code_ptr, target);
    case R_PPC_ADDR16:
        /*
         * We are (slightly) abusing this relocation type.  In particular,
         * assert that the low 2 bits are zero, and do not modify them.
         * That way we can use this with LD et al that have opcode bits
         * in the low 2 bits of the insn.
         */
        if ((value & 3) || value != (int16_t)value) {
            return false;
        }
        *code_ptr = (*code_ptr & ~0xfffc) | (value & 0xfffc);
        break;
    case R_PPC_ADDR32:
        /*
         * We are abusing this relocation type.  Again, this points to
         * a pair of insns, lis + load.  This is an absolute address
         * relocation for PPC32 so the lis cannot be removed.
         */
        lo = value;
        hi = value - lo;
        if (hi + lo != value) {
            return false;
        }
        code_ptr[0] = deposit32(code_ptr[0], 0, 16, hi >> 16);
        code_ptr[1] = deposit32(code_ptr[1], 0, 16, lo);
        break;
    default:
        g_assert_not_reached();
    }
    return true;
}

/* Ensure that the prefixed instruction does not cross a 64-byte boundary. */
static bool tcg_out_need_prefix_align(TCGContext *s)
{
    return ((uintptr_t)s->code_ptr & 0x3f) == 0x3c;
}

static void tcg_out_prefix_align(TCGContext *s)
{
    if (tcg_out_need_prefix_align(s)) {
        tcg_out32(s, NOP);
    }
}

static ptrdiff_t tcg_pcrel_diff_for_prefix(TCGContext *s, const void *target)
{
    return tcg_pcrel_diff(s, target) - (tcg_out_need_prefix_align(s) ? 4 : 0);
}

/* Output Type 00 Prefix - 8-Byte Load/Store Form (8LS:D) */
static void tcg_out_8ls_d(TCGContext *s, tcg_insn_unit opc, unsigned rt,
                          unsigned ra, tcg_target_long imm, bool r)
{
    tcg_insn_unit p, i;

    p = OPCD(1) | (r << 20) | ((imm >> 16) & 0x3ffff);
    i = opc | TAI(rt, ra, imm);

    tcg_out_prefix_align(s);
    tcg_out32(s, p);
    tcg_out32(s, i);
}

/* Output Type 10 Prefix - Modified Load/Store Form (MLS:D) */
static void tcg_out_mls_d(TCGContext *s, tcg_insn_unit opc, unsigned rt,
                          unsigned ra, tcg_target_long imm, bool r)
{
    tcg_insn_unit p, i;

    p = OPCD(1) | (2 << 24) | (r << 20) | ((imm >> 16) & 0x3ffff);
    i = opc | TAI(rt, ra, imm);

    tcg_out_prefix_align(s);
    tcg_out32(s, p);
    tcg_out32(s, i);
}

static void tcg_out_mem_long(TCGContext *s, int opi, int opx, TCGReg rt,
                             TCGReg base, tcg_target_long offset);

static bool tcg_out_mov(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg)
{
    if (ret == arg) {
        return true;
    }
    switch (type) {
    case TCG_TYPE_I64:
        tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
        /* fallthru */
    case TCG_TYPE_I32:
        if (ret < TCG_REG_V0) {
            if (arg < TCG_REG_V0) {
                tcg_out32(s, OR | SAB(arg, ret, arg));
                break;
            } else if (have_isa_2_07) {
                tcg_out32(s, (type == TCG_TYPE_I32 ? MFVSRWZ : MFVSRD)
                          | VRT(arg) | RA(ret));
                break;
            } else {
                /* Altivec does not support vector->integer moves.  */
                return false;
            }
        } else if (arg < TCG_REG_V0) {
            if (have_isa_2_07) {
                tcg_out32(s, (type == TCG_TYPE_I32 ? MTVSRWZ : MTVSRD)
                          | VRT(ret) | RA(arg));
                break;
            } else {
                /* Altivec does not support integer->vector moves.  */
                return false;
            }
        }
        /* fallthru */
    case TCG_TYPE_V64:
    case TCG_TYPE_V128:
        tcg_debug_assert(ret >= TCG_REG_V0 && arg >= TCG_REG_V0);
        tcg_out32(s, VOR | VRT(ret) | VRA(arg) | VRB(arg));
        break;
    default:
        g_assert_not_reached();
    }
    return true;
}

static void tcg_out_rld_rc(TCGContext *s, int op, TCGReg ra, TCGReg rs,
                           int sh, int mb, bool rc)
{
    tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
    sh = SH(sh & 0x1f) | (((sh >> 5) & 1) << 1);
    mb = MB64((mb >> 5) | ((mb << 1) & 0x3f));
    tcg_out32(s, op | RA(ra) | RS(rs) | sh | mb | rc);
}

static void tcg_out_rld(TCGContext *s, int op, TCGReg ra, TCGReg rs,
                        int sh, int mb)
{
    tcg_out_rld_rc(s, op, ra, rs, sh, mb, false);
}

static void tcg_out_rlw_rc(TCGContext *s, int op, TCGReg ra, TCGReg rs,
                           int sh, int mb, int me, bool rc)
{
    tcg_debug_assert((mb & 0x1f) == mb);
    tcg_debug_assert((me & 0x1f) == me);
    tcg_out32(s, op | RA(ra) | RS(rs) | SH(sh & 0x1f) | MB(mb) | ME(me) | rc);
}

static void tcg_out_rlw(TCGContext *s, int op, TCGReg ra, TCGReg rs,
                        int sh, int mb, int me)
{
    tcg_out_rlw_rc(s, op, ra, rs, sh, mb, me, false);
}

static void tcg_out_ext8s(TCGContext *s, TCGType type, TCGReg dst, TCGReg src)
{
    tcg_out32(s, EXTSB | RA(dst) | RS(src));
}

static void tcg_out_ext8u(TCGContext *s, TCGReg dst, TCGReg src)
{
    tcg_out32(s, ANDI | SAI(src, dst, 0xff));
}

static void tcg_out_ext16s(TCGContext *s, TCGType type, TCGReg dst, TCGReg src)
{
    tcg_out32(s, EXTSH | RA(dst) | RS(src));
}

static void tcg_out_ext16u(TCGContext *s, TCGReg dst, TCGReg src)
{
    tcg_out32(s, ANDI | SAI(src, dst, 0xffff));
}

static void tcg_out_ext32s(TCGContext *s, TCGReg dst, TCGReg src)
{
    tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
    tcg_out32(s, EXTSW | RA(dst) | RS(src));
}

static void tcg_out_ext32u(TCGContext *s, TCGReg dst, TCGReg src)
{
    tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
    tcg_out_rld(s, RLDICL, dst, src, 0, 32);
}

static void tcg_out_exts_i32_i64(TCGContext *s, TCGReg dst, TCGReg src)
{
    tcg_out_ext32s(s, dst, src);
}

static void tcg_out_extu_i32_i64(TCGContext *s, TCGReg dst, TCGReg src)
{
    tcg_out_ext32u(s, dst, src);
}

static void tcg_out_extrl_i64_i32(TCGContext *s, TCGReg rd, TCGReg rn)
{
    tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
    tcg_out_mov(s, TCG_TYPE_I32, rd, rn);
}

static inline void tcg_out_shli32(TCGContext *s, TCGReg dst, TCGReg src, int c)
{
    tcg_out_rlw(s, RLWINM, dst, src, c, 0, 31 - c);
}

static inline void tcg_out_shli64(TCGContext *s, TCGReg dst, TCGReg src, int c)
{
    tcg_out_rld(s, RLDICR, dst, src, c, 63 - c);
}

static inline void tcg_out_sari32(TCGContext *s, TCGReg dst, TCGReg src, int c)
{
    /* Limit immediate shift count lest we create an illegal insn.  */
    tcg_out32(s, SRAWI | RA(dst) | RS(src) | SH(c & 31));
}

static inline void tcg_out_shri32(TCGContext *s, TCGReg dst, TCGReg src, int c)
{
    tcg_out_rlw(s, RLWINM, dst, src, 32 - c, c, 31);
}

static inline void tcg_out_shri64(TCGContext *s, TCGReg dst, TCGReg src, int c)
{
    tcg_out_rld(s, RLDICL, dst, src, 64 - c, c);
}

static inline void tcg_out_sari64(TCGContext *s, TCGReg dst, TCGReg src, int c)
{
    tcg_out32(s, SRADI | RA(dst) | RS(src) | SH(c & 0x1f) | ((c >> 4) & 2));
}

static void tcg_out_addpcis(TCGContext *s, TCGReg dst, intptr_t imm)
{
    uint32_t d0, d1, d2;

    tcg_debug_assert((imm & 0xffff) == 0);
    tcg_debug_assert(imm == (int32_t)imm);

    d2 = extract32(imm, 16, 1);
    d1 = extract32(imm, 17, 5);
    d0 = extract32(imm, 22, 10);
    tcg_out32(s, ADDPCIS | RT(dst) | (d1 << 16) | (d0 << 6) | d2);
}

static void tcg_out_bswap32(TCGContext *s, TCGReg dst, TCGReg src, int flags)
{
    TCGReg tmp = dst == src ? TCG_REG_R0 : dst;

    if (have_isa_3_10) {
        tcg_out32(s, BRW | RA(dst) | RS(src));
        if (flags & TCG_BSWAP_OS) {
            tcg_out_ext32s(s, dst, dst);
        } else if ((flags & (TCG_BSWAP_IZ | TCG_BSWAP_OZ)) == TCG_BSWAP_OZ) {
            tcg_out_ext32u(s, dst, dst);
        }
        return;
    }

    /*
     * Stolen from gcc's builtin_bswap32.
     * In the following,
     *   dep(a, b, m) -> (a & ~m) | (b & m)
     *
     * Begin with:                              src = xxxxabcd
     */
    /* tmp = rol32(src, 8) & 0xffffffff             = 0000bcda */
    tcg_out_rlw(s, RLWINM, tmp, src, 8, 0, 31);
    /* tmp = dep(tmp, rol32(src, 24), 0xff000000)   = 0000dcda */
    tcg_out_rlw(s, RLWIMI, tmp, src, 24, 0, 7);
    /* tmp = dep(tmp, rol32(src, 24), 0x0000ff00)   = 0000dcba */
    tcg_out_rlw(s, RLWIMI, tmp, src, 24, 16, 23);

    if (flags & TCG_BSWAP_OS) {
        tcg_out_ext32s(s, dst, tmp);
    } else {
        tcg_out_mov(s, TCG_TYPE_REG, dst, tmp);
    }
}

static void tcg_out_bswap64(TCGContext *s, TCGReg dst, TCGReg src)
{
    TCGReg t0 = dst == src ? TCG_REG_R0 : dst;
    TCGReg t1 = dst == src ? dst : TCG_REG_R0;

    if (have_isa_3_10) {
        tcg_out32(s, BRD | RA(dst) | RS(src));
        return;
    }

    /*
     * In the following,
     *   dep(a, b, m) -> (a & ~m) | (b & m)
     *
     * Begin with:                              src = abcdefgh
     */
    /* t0 = rol32(src, 8) & 0xffffffff              = 0000fghe */
    tcg_out_rlw(s, RLWINM, t0, src, 8, 0, 31);
    /* t0 = dep(t0, rol32(src, 24), 0xff000000)     = 0000hghe */
    tcg_out_rlw(s, RLWIMI, t0, src, 24, 0, 7);
    /* t0 = dep(t0, rol32(src, 24), 0x0000ff00)     = 0000hgfe */
    tcg_out_rlw(s, RLWIMI, t0, src, 24, 16, 23);

    /* t0 = rol64(t0, 32)                           = hgfe0000 */
    tcg_out_rld(s, RLDICL, t0, t0, 32, 0);
    /* t1 = rol64(src, 32)                          = efghabcd */
    tcg_out_rld(s, RLDICL, t1, src, 32, 0);

    /* t0 = dep(t0, rol32(t1, 24), 0xffffffff)      = hgfebcda */
    tcg_out_rlw(s, RLWIMI, t0, t1, 8, 0, 31);
    /* t0 = dep(t0, rol32(t1, 24), 0xff000000)      = hgfedcda */
    tcg_out_rlw(s, RLWIMI, t0, t1, 24, 0, 7);
    /* t0 = dep(t0, rol32(t1, 24), 0x0000ff00)      = hgfedcba */
    tcg_out_rlw(s, RLWIMI, t0, t1, 24, 16, 23);

    tcg_out_mov(s, TCG_TYPE_REG, dst, t0);
}

/* Emit a move into ret of arg, if it can be done in one insn.  */
static bool tcg_out_movi_one(TCGContext *s, TCGReg ret, tcg_target_long arg)
{
    if (arg == (int16_t)arg) {
        tcg_out32(s, ADDI | TAI(ret, 0, arg));
        return true;
    }
    if (arg == (int32_t)arg && (arg & 0xffff) == 0) {
        tcg_out32(s, ADDIS | TAI(ret, 0, arg >> 16));
        return true;
    }
    return false;
}

static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret,
                             tcg_target_long arg, bool in_prologue)
{
    intptr_t tb_diff;
    tcg_target_long tmp;
    int shift;

    tcg_debug_assert(TCG_TARGET_REG_BITS == 64 || type == TCG_TYPE_I32);

    if (TCG_TARGET_REG_BITS == 64 && type == TCG_TYPE_I32) {
        arg = (int32_t)arg;
    }

    /* Load 16-bit immediates with one insn.  */
    if (tcg_out_movi_one(s, ret, arg)) {
        return;
    }

    /* Load addresses within the TB with one insn.  */
    tb_diff = ppc_tbrel_diff(s, (void *)arg);
    if (!in_prologue && USE_REG_TB && tb_diff == (int16_t)tb_diff) {
        tcg_out32(s, ADDI | TAI(ret, TCG_REG_TB, tb_diff));
        return;
    }

    /*
     * Load values up to 34 bits, and pc-relative addresses,
     * with one prefixed insn.
     */
    if (have_isa_3_10) {
        if (arg == sextract64(arg, 0, 34)) {
            /* pli ret,value = paddi ret,0,value,0 */
            tcg_out_mls_d(s, ADDI, ret, 0, arg, 0);
            return;
        }

        tmp = tcg_pcrel_diff_for_prefix(s, (void *)arg);
        if (tmp == sextract64(tmp, 0, 34)) {
            /* pla ret,value = paddi ret,0,value,1 */
            tcg_out_mls_d(s, ADDI, ret, 0, tmp, 1);
            return;
        }
    }

    /* Load 32-bit immediates with two insns.  Note that we've already
       eliminated bare ADDIS, so we know both insns are required.  */
    if (TCG_TARGET_REG_BITS == 32 || arg == (int32_t)arg) {
        tcg_out32(s, ADDIS | TAI(ret, 0, arg >> 16));
        tcg_out32(s, ORI | SAI(ret, ret, arg));
        return;
    }
    if (arg == (uint32_t)arg && !(arg & 0x8000)) {
        tcg_out32(s, ADDI | TAI(ret, 0, arg));
        tcg_out32(s, ORIS | SAI(ret, ret, arg >> 16));
        return;
    }

    /* Load masked 16-bit value.  */
    if (arg > 0 && (arg & 0x8000)) {
        tmp = arg | 0x7fff;
        if ((tmp & (tmp + 1)) == 0) {
            int mb = clz64(tmp + 1) + 1;
            tcg_out32(s, ADDI | TAI(ret, 0, arg));
            tcg_out_rld(s, RLDICL, ret, ret, 0, mb);
            return;
        }
    }

    /* Load common masks with 2 insns.  */
    shift = ctz64(arg);
    tmp = arg >> shift;
    if (tmp == (int16_t)tmp) {
        tcg_out32(s, ADDI | TAI(ret, 0, tmp));
        tcg_out_shli64(s, ret, ret, shift);
        return;
    }
    shift = clz64(arg);
    if (tcg_out_movi_one(s, ret, arg << shift)) {
        tcg_out_shri64(s, ret, ret, shift);
        return;
    }

    /* Load addresses within 2GB with 2 insns. */
    if (have_isa_3_00) {
        intptr_t hi = tcg_pcrel_diff(s, (void *)arg) - 4;
        int16_t lo = hi;

        hi -= lo;
        if (hi == (int32_t)hi) {
            tcg_out_addpcis(s, TCG_REG_TMP2, hi);
            tcg_out32(s, ADDI | TAI(ret, TCG_REG_TMP2, lo));
            return;
        }
    }

    /* Load addresses within 2GB of TB with 2 (or rarely 3) insns.  */
    if (!in_prologue && USE_REG_TB && tb_diff == (int32_t)tb_diff) {
        tcg_out_mem_long(s, ADDI, ADD, ret, TCG_REG_TB, tb_diff);
        return;
    }

    /* Use the constant pool, if possible.  */
    if (!in_prologue && USE_REG_TB) {
        new_pool_label(s, arg, R_PPC_ADDR16, s->code_ptr,
                       ppc_tbrel_diff(s, NULL));
        tcg_out32(s, LD | TAI(ret, TCG_REG_TB, 0));
        return;
    }
    if (have_isa_3_10) {
        tcg_out_8ls_d(s, PLD, ret, 0, 0, 1);
        new_pool_label(s, arg, R_PPC64_PCREL34, s->code_ptr - 2, 0);
        return;
    }
    if (have_isa_3_00) {
        tcg_out_addpcis(s, TCG_REG_TMP2, 0);
        new_pool_label(s, arg, R_PPC_REL14, s->code_ptr, 0);
        tcg_out32(s, LD | TAI(ret, TCG_REG_TMP2, 0));
        return;
    }

    tmp = arg >> 31 >> 1;
    tcg_out_movi(s, TCG_TYPE_I32, ret, tmp);
    if (tmp) {
        tcg_out_shli64(s, ret, ret, 32);
    }
    if (arg & 0xffff0000) {
        tcg_out32(s, ORIS | SAI(ret, ret, arg >> 16));
    }
    if (arg & 0xffff) {
        tcg_out32(s, ORI | SAI(ret, ret, arg));
    }
}

static void tcg_out_dupi_vec(TCGContext *s, TCGType type, unsigned vece,
                             TCGReg ret, int64_t val)
{
    uint32_t load_insn;
    int rel, low;
    intptr_t add;

    switch (vece) {
    case MO_8:
        low = (int8_t)val;
        if (low >= -16 && low < 16) {
            tcg_out32(s, VSPLTISB | VRT(ret) | ((val & 31) << 16));
            return;
        }
        if (have_isa_3_00) {
            tcg_out32(s, XXSPLTIB | VRT(ret) | ((val & 0xff) << 11));
            return;
        }
        break;

    case MO_16:
        low = (int16_t)val;
        if (low >= -16 && low < 16) {
            tcg_out32(s, VSPLTISH | VRT(ret) | ((val & 31) << 16));
            return;
        }
        break;

    case MO_32:
        low = (int32_t)val;
        if (low >= -16 && low < 16) {
            tcg_out32(s, VSPLTISW | VRT(ret) | ((val & 31) << 16));
            return;
        }
        break;
    }

    /*
     * Otherwise we must load the value from the constant pool.
     */
    if (USE_REG_TB) {
        rel = R_PPC_ADDR16;
        add = ppc_tbrel_diff(s, NULL);
    } else if (have_isa_3_10) {
        if (type == TCG_TYPE_V64) {
            tcg_out_8ls_d(s, PLXSD, ret & 31, 0, 0, 1);
            new_pool_label(s, val, R_PPC64_PCREL34, s->code_ptr - 2, 0);
        } else {
            tcg_out_8ls_d(s, PLXV, ret & 31, 0, 0, 1);
            new_pool_l2(s, R_PPC64_PCREL34, s->code_ptr - 2, 0, val, val);
        }
        return;
    } else if (have_isa_3_00) {
        tcg_out_addpcis(s, TCG_REG_TMP1, 0);
        rel = R_PPC_REL14;
        add = 0;
    } else {
        rel = R_PPC_ADDR32;
        add = 0;
    }

    if (have_vsx) {
        load_insn = type == TCG_TYPE_V64 ? LXSDX : LXVDSX;
        load_insn |= VRT(ret) | RB(TCG_REG_TMP1);
        if (TCG_TARGET_REG_BITS == 64) {
            new_pool_label(s, val, rel, s->code_ptr, add);
        } else {
            new_pool_l2(s, rel, s->code_ptr, add, val >> 32, val);
        }
    } else {
        load_insn = LVX | VRT(ret) | RB(TCG_REG_TMP1);
        if (TCG_TARGET_REG_BITS == 64) {
            new_pool_l2(s, rel, s->code_ptr, add, val, val);
        } else {
            new_pool_l4(s, rel, s->code_ptr, add,
                        val >> 32, val, val >> 32, val);
        }
    }

    if (USE_REG_TB) {
        tcg_out32(s, ADDI | TAI(TCG_REG_TMP1, 0, 0));
        load_insn |= RA(TCG_REG_TB);
    } else if (have_isa_3_00) {
        tcg_out32(s, ADDI | TAI(TCG_REG_TMP1, TCG_REG_TMP1, 0));
    } else {
        tcg_out32(s, ADDIS | TAI(TCG_REG_TMP1, 0, 0));
        tcg_out32(s, ADDI | TAI(TCG_REG_TMP1, TCG_REG_TMP1, 0));
    }
    tcg_out32(s, load_insn);
}

static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg ret,
                         tcg_target_long arg)
{
    switch (type) {
    case TCG_TYPE_I32:
    case TCG_TYPE_I64:
        tcg_debug_assert(ret < TCG_REG_V0);
        tcg_out_movi_int(s, type, ret, arg, false);
        break;

    default:
        g_assert_not_reached();
    }
}

static bool tcg_out_xchg(TCGContext *s, TCGType type, TCGReg r1, TCGReg r2)
{
    return false;
}

static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
                             tcg_target_long imm)
{
    /* This function is only used for passing structs by reference. */
    g_assert_not_reached();
}

static bool mask_operand(uint32_t c, int *mb, int *me)
{
    uint32_t lsb, test;

    /* Accept a bit pattern like:
           0....01....1
           1....10....0
           0..01..10..0
       Keep track of the transitions.  */
    if (c == 0 || c == -1) {
        return false;
    }
    test = c;
    lsb = test & -test;
    test += lsb;
    if (test & (test - 1)) {
        return false;
    }

    *me = clz32(lsb);
    *mb = test ? clz32(test & -test) + 1 : 0;
    return true;
}

static bool mask64_operand(uint64_t c, int *mb, int *me)
{
    uint64_t lsb;

    if (c == 0) {
        return false;
    }

    lsb = c & -c;
    /* Accept 1..10..0.  */
    if (c == -lsb) {
        *mb = 0;
        *me = clz64(lsb);
        return true;
    }
    /* Accept 0..01..1.  */
    if (lsb == 1 && (c & (c + 1)) == 0) {
        *mb = clz64(c + 1) + 1;
        *me = 63;
        return true;
    }
    return false;
}

static void tcg_out_andi32(TCGContext *s, TCGReg dst, TCGReg src, uint32_t c)
{
    int mb, me;

    if (mask_operand(c, &mb, &me)) {
        tcg_out_rlw(s, RLWINM, dst, src, 0, mb, me);
    } else if ((c & 0xffff) == c) {
        tcg_out32(s, ANDI | SAI(src, dst, c));
        return;
    } else if ((c & 0xffff0000) == c) {
        tcg_out32(s, ANDIS | SAI(src, dst, c >> 16));
        return;
    } else {
        tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_R0, c);
        tcg_out32(s, AND | SAB(src, dst, TCG_REG_R0));
    }
}

static void tcg_out_andi64(TCGContext *s, TCGReg dst, TCGReg src, uint64_t c)
{
    int mb, me;

    tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
    if (mask64_operand(c, &mb, &me)) {
        if (mb == 0) {
            tcg_out_rld(s, RLDICR, dst, src, 0, me);
        } else {
            tcg_out_rld(s, RLDICL, dst, src, 0, mb);
        }
    } else if ((c & 0xffff) == c) {
        tcg_out32(s, ANDI | SAI(src, dst, c));
        return;
    } else if ((c & 0xffff0000) == c) {
        tcg_out32(s, ANDIS | SAI(src, dst, c >> 16));
        return;
    } else {
        tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_R0, c);
        tcg_out32(s, AND | SAB(src, dst, TCG_REG_R0));
    }
}

static void tcg_out_zori32(TCGContext *s, TCGReg dst, TCGReg src, uint32_t c,
                           int op_lo, int op_hi)
{
    if (c >> 16) {
        tcg_out32(s, op_hi | SAI(src, dst, c >> 16));
        src = dst;
    }
    if (c & 0xffff) {
        tcg_out32(s, op_lo | SAI(src, dst, c));
        src = dst;
    }
}

static void tcg_out_ori32(TCGContext *s, TCGReg dst, TCGReg src, uint32_t c)
{
    tcg_out_zori32(s, dst, src, c, ORI, ORIS);
}

static void tcg_out_xori32(TCGContext *s, TCGReg dst, TCGReg src, uint32_t c)
{
    tcg_out_zori32(s, dst, src, c, XORI, XORIS);
}

static void tcg_out_b(TCGContext *s, int mask, const tcg_insn_unit *target)
{
    ptrdiff_t disp = tcg_pcrel_diff(s, target);
    if (in_range_b(disp)) {
        tcg_out32(s, B | (disp & 0x3fffffc) | mask);
    } else {
        tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R0, (uintptr_t)target);
        tcg_out32(s, MTSPR | RS(TCG_REG_R0) | CTR);
        tcg_out32(s, BCCTR | BO_ALWAYS | mask);
    }
}

static void tcg_out_mem_long(TCGContext *s, int opi, int opx, TCGReg rt,
                             TCGReg base, tcg_target_long offset)
{
    tcg_target_long orig = offset, l0, l1, extra = 0, align = 0;
    bool is_int_store = false;
    TCGReg rs = TCG_REG_TMP1;

    switch (opi) {
    case LD: case LWA:
        align = 3;
        /* FALLTHRU */
    default:
        if (rt > TCG_REG_R0 && rt < TCG_REG_V0) {
            rs = rt;
            break;
        }
        break;
    case LXSD:
    case STXSD:
        align = 3;
        break;
    case LXV:
    case STXV:
        align = 15;
        break;
    case STD:
        align = 3;
        /* FALLTHRU */
    case STB: case STH: case STW:
        is_int_store = true;
        break;
    }

    /* For unaligned or large offsets, use the prefixed form. */
    if (have_isa_3_10
        && (offset != (int16_t)offset || (offset & align))
        && offset == sextract64(offset, 0, 34)) {
        /*
         * Note that the MLS:D insns retain their un-prefixed opcode,
         * while the 8LS:D insns use a different opcode space.
         */
        switch (opi) {
        case LBZ:
        case LHZ:
        case LHA:
        case LWZ:
        case STB:
        case STH:
        case STW:
        case ADDI:
            tcg_out_mls_d(s, opi, rt, base, offset, 0);
            return;
        case LWA:
            tcg_out_8ls_d(s, PLWA, rt, base, offset, 0);
            return;
        case LD:
            tcg_out_8ls_d(s, PLD, rt, base, offset, 0);
            return;
        case STD:
            tcg_out_8ls_d(s, PSTD, rt, base, offset, 0);
            return;
        case LXSD:
            tcg_out_8ls_d(s, PLXSD, rt & 31, base, offset, 0);
            return;
        case STXSD:
            tcg_out_8ls_d(s, PSTXSD, rt & 31, base, offset, 0);
            return;
        case LXV:
            tcg_out_8ls_d(s, PLXV, rt & 31, base, offset, 0);
            return;
        case STXV:
            tcg_out_8ls_d(s, PSTXV, rt & 31, base, offset, 0);
            return;
        }
    }

    /* For unaligned, or very large offsets, use the indexed form.  */
    if (offset & align || offset != (int32_t)offset || opi == 0) {
        if (rs == base) {
            rs = TCG_REG_R0;
        }
        tcg_debug_assert(!is_int_store || rs != rt);
        tcg_out_movi(s, TCG_TYPE_PTR, rs, orig);
        tcg_out32(s, opx | TAB(rt & 31, base, rs));
        return;
    }

    l0 = (int16_t)offset;
    offset = (offset - l0) >> 16;
    l1 = (int16_t)offset;

    if (l1 < 0 && orig >= 0) {
        extra = 0x4000;
        l1 = (int16_t)(offset - 0x4000);
    }
    if (l1) {
        tcg_out32(s, ADDIS | TAI(rs, base, l1));
        base = rs;
    }
    if (extra) {
        tcg_out32(s, ADDIS | TAI(rs, base, extra));
        base = rs;
    }
    if (opi != ADDI || base != rt || l0 != 0) {
        tcg_out32(s, opi | TAI(rt & 31, base, l0));
    }
}

static void tcg_out_vsldoi(TCGContext *s, TCGReg ret,
                           TCGReg va, TCGReg vb, int shb)
{
    tcg_out32(s, VSLDOI | VRT(ret) | VRA(va) | VRB(vb) | (shb << 6));
}

static void tcg_out_ld(TCGContext *s, TCGType type, TCGReg ret,
                       TCGReg base, intptr_t offset)
{
    int shift;

    switch (type) {
    case TCG_TYPE_I32:
        if (ret < TCG_REG_V0) {
            tcg_out_mem_long(s, LWZ, LWZX, ret, base, offset);
            break;
        }
        if (have_isa_2_07 && have_vsx) {
            tcg_out_mem_long(s, 0, LXSIWZX, ret, base, offset);
            break;
        }
        tcg_debug_assert((offset & 3) == 0);
        tcg_out_mem_long(s, 0, LVEWX, ret, base, offset);
        shift = (offset - 4) & 0xc;
        if (shift) {
            tcg_out_vsldoi(s, ret, ret, ret, shift);
        }
        break;
    case TCG_TYPE_I64:
        if (ret < TCG_REG_V0) {
            tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
            tcg_out_mem_long(s, LD, LDX, ret, base, offset);
            break;
        }
        /* fallthru */
    case TCG_TYPE_V64:
        tcg_debug_assert(ret >= TCG_REG_V0);
        if (have_vsx) {
            tcg_out_mem_long(s, have_isa_3_00 ? LXSD : 0, LXSDX,
                             ret, base, offset);
            break;
        }
        tcg_debug_assert((offset & 7) == 0);
        tcg_out_mem_long(s, 0, LVX, ret, base, offset & -16);
        if (offset & 8) {
            tcg_out_vsldoi(s, ret, ret, ret, 8);
        }
        break;
    case TCG_TYPE_V128:
        tcg_debug_assert(ret >= TCG_REG_V0);
        tcg_debug_assert((offset & 15) == 0);
        tcg_out_mem_long(s, have_isa_3_00 ? LXV : 0,
                         LVX, ret, base, offset);
        break;
    default:
        g_assert_not_reached();
    }
}

static void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
                              TCGReg base, intptr_t offset)
{
    int shift;

    switch (type) {
    case TCG_TYPE_I32:
        if (arg < TCG_REG_V0) {
            tcg_out_mem_long(s, STW, STWX, arg, base, offset);
            break;
        }
        if (have_isa_2_07 && have_vsx) {
            tcg_out_mem_long(s, 0, STXSIWX, arg, base, offset);
            break;
        }
        assert((offset & 3) == 0);
        tcg_debug_assert((offset & 3) == 0);
        shift = (offset - 4) & 0xc;
        if (shift) {
            tcg_out_vsldoi(s, TCG_VEC_TMP1, arg, arg, shift);
            arg = TCG_VEC_TMP1;
        }
        tcg_out_mem_long(s, 0, STVEWX, arg, base, offset);
        break;
    case TCG_TYPE_I64:
        if (arg < TCG_REG_V0) {
            tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
            tcg_out_mem_long(s, STD, STDX, arg, base, offset);
            break;
        }
        /* fallthru */
    case TCG_TYPE_V64:
        tcg_debug_assert(arg >= TCG_REG_V0);
        if (have_vsx) {
            tcg_out_mem_long(s, have_isa_3_00 ? STXSD : 0,
                             STXSDX, arg, base, offset);
            break;
        }
        tcg_debug_assert((offset & 7) == 0);
        if (offset & 8) {
            tcg_out_vsldoi(s, TCG_VEC_TMP1, arg, arg, 8);
            arg = TCG_VEC_TMP1;
        }
        tcg_out_mem_long(s, 0, STVEWX, arg, base, offset);
        tcg_out_mem_long(s, 0, STVEWX, arg, base, offset + 4);
        break;
    case TCG_TYPE_V128:
        tcg_debug_assert(arg >= TCG_REG_V0);
        tcg_out_mem_long(s, have_isa_3_00 ? STXV : 0,
                         STVX, arg, base, offset);
        break;
    default:
        g_assert_not_reached();
    }
}

static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
                               TCGReg base, intptr_t ofs)
{
    return false;
}

/*
 * Set dest non-zero if and only if (arg1 & arg2) is non-zero.
 * If RC, then also set RC0.
 */
static void tcg_out_test(TCGContext *s, TCGReg dest, TCGReg arg1, TCGArg arg2,
                         bool const_arg2, TCGType type, bool rc)
{
    int mb, me;

    if (!const_arg2) {
        tcg_out32(s, AND | SAB(arg1, dest, arg2) | rc);
        return;
    }

    if (type == TCG_TYPE_I32) {
        arg2 = (uint32_t)arg2;
    }

    if ((arg2 & ~0xffff) == 0) {
        tcg_out32(s, ANDI | SAI(arg1, dest, arg2));
        return;
    }
    if ((arg2 & ~0xffff0000ull) == 0) {
        tcg_out32(s, ANDIS | SAI(arg1, dest, arg2 >> 16));
        return;
    }
    if (arg2 == (uint32_t)arg2 && mask_operand(arg2, &mb, &me)) {
        tcg_out_rlw_rc(s, RLWINM, dest, arg1, 0, mb, me, rc);
        return;
    }
    if (TCG_TARGET_REG_BITS == 64) {
        int sh = clz64(arg2);
        if (mask64_operand(arg2 << sh, &mb, &me)) {
            tcg_out_rld_rc(s, RLDICR, dest, arg1, sh, me, rc);
            return;
        }
    }
    /* Constraints should satisfy this. */
    g_assert_not_reached();
}

static void tcg_out_cmp(TCGContext *s, int cond, TCGArg arg1, TCGArg arg2,
                        bool const_arg2, int cr, TCGType type)
{
    uint32_t op;

    tcg_debug_assert(TCG_TARGET_REG_BITS == 64 || type == TCG_TYPE_I32);

    /*
     * Simplify the comparisons below wrt CMPI.
     * All of the tests are 16-bit, so a 32-bit sign extend always works.
     */
    if (type == TCG_TYPE_I32) {
        arg2 = (int32_t)arg2;
    }

    switch (cond) {
    case TCG_COND_EQ:
    case TCG_COND_NE:
        if (const_arg2) {
            if ((int16_t)arg2 == arg2) {
                op = CMPI;
                break;
            }
            tcg_debug_assert((uint16_t)arg2 == arg2);
            op = CMPLI;
            break;
        }
        op = CMPL;
        break;

    case TCG_COND_TSTEQ:
    case TCG_COND_TSTNE:
        tcg_debug_assert(cr == 0);
        tcg_out_test(s, TCG_REG_R0, arg1, arg2, const_arg2, type, true);
        return;

    case TCG_COND_LT:
    case TCG_COND_GE:
    case TCG_COND_LE:
    case TCG_COND_GT:
        if (const_arg2) {
            tcg_debug_assert((int16_t)arg2 == arg2);
            op = CMPI;
            break;
        }
        op = CMP;
        break;

    case TCG_COND_LTU:
    case TCG_COND_GEU:
    case TCG_COND_LEU:
    case TCG_COND_GTU:
        if (const_arg2) {
            tcg_debug_assert((uint16_t)arg2 == arg2);
            op = CMPLI;
            break;
        }
        op = CMPL;
        break;

    default:
        g_assert_not_reached();
    }
    op |= BF(cr) | ((type == TCG_TYPE_I64) << 21);
    op |= RA(arg1);
    op |= const_arg2 ? arg2 & 0xffff : RB(arg2);
    tcg_out32(s, op);
}

static void tcg_out_setcond_eq0(TCGContext *s, TCGType type,
                                TCGReg dst, TCGReg src, bool neg)
{
    if (neg && (TCG_TARGET_REG_BITS == 32 || type == TCG_TYPE_I64)) {
        /*
         * X != 0 implies X + -1 generates a carry.
         * RT = (~X + X) + CA
         *    = -1 + CA
         *    = CA ? 0 : -1
         */
        tcg_out32(s, ADDIC | TAI(TCG_REG_R0, src, -1));
        tcg_out32(s, SUBFE | TAB(dst, src, src));
        return;
    }

    if (type == TCG_TYPE_I32) {
        tcg_out32(s, CNTLZW | RS(src) | RA(dst));
        tcg_out_shri32(s, dst, dst, 5);
    } else {
        tcg_out32(s, CNTLZD | RS(src) | RA(dst));
        tcg_out_shri64(s, dst, dst, 6);
    }
    if (neg) {
        tcg_out32(s, NEG | RT(dst) | RA(dst));
    }
}

static void tcg_out_setcond_ne0(TCGContext *s, TCGType type,
                                TCGReg dst, TCGReg src, bool neg)
{
    if (!neg && (TCG_TARGET_REG_BITS == 32 || type == TCG_TYPE_I64)) {
        /*
         * X != 0 implies X + -1 generates a carry.  Extra addition
         * trickery means: R = X-1 + ~X + C = X-1 + (-X+1) + C = C.
         */
        tcg_out32(s, ADDIC | TAI(TCG_REG_R0, src, -1));
        tcg_out32(s, SUBFE | TAB(dst, TCG_REG_R0, src));
        return;
    }
    tcg_out_setcond_eq0(s, type, dst, src, false);
    if (neg) {
        tcg_out32(s, ADDI | TAI(dst, dst, -1));
    } else {
        tcg_out_xori32(s, dst, dst, 1);
    }
}

static TCGReg tcg_gen_setcond_xor(TCGContext *s, TCGReg arg1, TCGArg arg2,
                                  bool const_arg2)
{
    if (const_arg2) {
        if ((uint32_t)arg2 == arg2) {
            tcg_out_xori32(s, TCG_REG_R0, arg1, arg2);
        } else {
            tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_R0, arg2);
            tcg_out32(s, XOR | SAB(arg1, TCG_REG_R0, TCG_REG_R0));
        }
    } else {
        tcg_out32(s, XOR | SAB(arg1, TCG_REG_R0, arg2));
    }
    return TCG_REG_R0;
}

static void tcg_out_setcond(TCGContext *s, TCGType type, TCGCond cond,
                            TCGReg arg0, TCGReg arg1, TCGArg arg2,
                            bool const_arg2, bool neg)
{
    int sh;
    bool inv;

    tcg_debug_assert(TCG_TARGET_REG_BITS == 64 || type == TCG_TYPE_I32);

    /* Ignore high bits of a potential constant arg2.  */
    if (type == TCG_TYPE_I32) {
        arg2 = (uint32_t)arg2;
    }

    /* With SETBC/SETBCR, we can always implement with 2 insns. */
    if (have_isa_3_10) {
        tcg_insn_unit bi, opc;

        tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 0, type);

        /* Re-use tcg_to_bc for BI and BO_COND_{TRUE,FALSE}. */
        bi = tcg_to_bc[cond] & (0x1f << 16);
        if (tcg_to_bc[cond] & BO(8)) {
            opc = neg ? SETNBC : SETBC;
        } else {
            opc = neg ? SETNBCR : SETBCR;
        }
        tcg_out32(s, opc | RT(arg0) | bi);
        return;
    }

    /* Handle common and trivial cases before handling anything else.  */
    if (arg2 == 0) {
        switch (cond) {
        case TCG_COND_EQ:
            tcg_out_setcond_eq0(s, type, arg0, arg1, neg);
            return;
        case TCG_COND_NE:
            tcg_out_setcond_ne0(s, type, arg0, arg1, neg);
            return;
        case TCG_COND_GE:
            tcg_out32(s, NOR | SAB(arg1, arg0, arg1));
            arg1 = arg0;
            /* FALLTHRU */
        case TCG_COND_LT:
            /* Extract the sign bit.  */
            if (type == TCG_TYPE_I32) {
                if (neg) {
                    tcg_out_sari32(s, arg0, arg1, 31);
                } else {
                    tcg_out_shri32(s, arg0, arg1, 31);
                }
            } else {
                if (neg) {
                    tcg_out_sari64(s, arg0, arg1, 63);
                } else {
                    tcg_out_shri64(s, arg0, arg1, 63);
                }
            }
            return;
        default:
            break;
        }
    }

    /* If we have ISEL, we can implement everything with 3 or 4 insns.
       All other cases below are also at least 3 insns, so speed up the
       code generator by not considering them and always using ISEL.  */
    if (have_isel) {
        int isel, tab;

        tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 0, type);

        isel = tcg_to_isel[cond];

        tcg_out_movi(s, type, arg0, neg ? -1 : 1);
        if (isel & 1) {
            /* arg0 = (bc ? 0 : 1) */
            tab = TAB(arg0, 0, arg0);
            isel &= ~1;
        } else {
            /* arg0 = (bc ? 1 : 0) */
            tcg_out_movi(s, type, TCG_REG_R0, 0);
            tab = TAB(arg0, arg0, TCG_REG_R0);
        }
        tcg_out32(s, isel | tab);
        return;
    }

    inv = false;
    switch (cond) {
    case TCG_COND_EQ:
        arg1 = tcg_gen_setcond_xor(s, arg1, arg2, const_arg2);
        tcg_out_setcond_eq0(s, type, arg0, arg1, neg);
        break;

    case TCG_COND_NE:
        arg1 = tcg_gen_setcond_xor(s, arg1, arg2, const_arg2);
        tcg_out_setcond_ne0(s, type, arg0, arg1, neg);
        break;

    case TCG_COND_TSTEQ:
        tcg_out_test(s, TCG_REG_R0, arg1, arg2, const_arg2, type, false);
        tcg_out_setcond_eq0(s, type, arg0, TCG_REG_R0, neg);
        break;

    case TCG_COND_TSTNE:
        tcg_out_test(s, TCG_REG_R0, arg1, arg2, const_arg2, type, false);
        tcg_out_setcond_ne0(s, type, arg0, TCG_REG_R0, neg);
        break;

    case TCG_COND_LE:
    case TCG_COND_LEU:
        inv = true;
        /* fall through */
    case TCG_COND_GT:
    case TCG_COND_GTU:
        sh = 30; /* CR7 CR_GT */
        goto crtest;

    case TCG_COND_GE:
    case TCG_COND_GEU:
        inv = true;
        /* fall through */
    case TCG_COND_LT:
    case TCG_COND_LTU:
        sh = 29; /* CR7 CR_LT */
        goto crtest;

    crtest:
        tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 7, type);
        tcg_out32(s, MFOCRF | RT(TCG_REG_R0) | FXM(7));
        tcg_out_rlw(s, RLWINM, arg0, TCG_REG_R0, sh, 31, 31);
        if (neg && inv) {
            tcg_out32(s, ADDI | TAI(arg0, arg0, -1));
        } else if (neg) {
            tcg_out32(s, NEG | RT(arg0) | RA(arg0));
        } else if (inv) {
            tcg_out_xori32(s, arg0, arg0, 1);
        }
        break;

    default:
        g_assert_not_reached();
    }
}

static void tgen_setcond(TCGContext *s, TCGType type, TCGCond cond,
                         TCGReg dest, TCGReg arg1, TCGReg arg2)
{
    tcg_out_setcond(s, type, cond, dest, arg1, arg2, false, false);
}

static void tgen_setcondi(TCGContext *s, TCGType type, TCGCond cond,
                          TCGReg dest, TCGReg arg1, tcg_target_long arg2)
{
    tcg_out_setcond(s, type, cond, dest, arg1, arg2, true, false);
}

static const TCGOutOpSetcond outop_setcond = {
    .base.static_constraint = C_O1_I2(r, r, rC),
    .out_rrr = tgen_setcond,
    .out_rri = tgen_setcondi,
};

static void tgen_negsetcond(TCGContext *s, TCGType type, TCGCond cond,
                            TCGReg dest, TCGReg arg1, TCGReg arg2)
{
    tcg_out_setcond(s, type, cond, dest, arg1, arg2, false, true);
}

static void tgen_negsetcondi(TCGContext *s, TCGType type, TCGCond cond,
                             TCGReg dest, TCGReg arg1, tcg_target_long arg2)
{
    tcg_out_setcond(s, type, cond, dest, arg1, arg2, true, true);
}

static const TCGOutOpSetcond outop_negsetcond = {
    .base.static_constraint = C_O1_I2(r, r, rC),
    .out_rrr = tgen_negsetcond,
    .out_rri = tgen_negsetcondi,
};

static void tcg_out_bc(TCGContext *s, TCGCond cond, int bd)
{
    tcg_out32(s, tcg_to_bc[cond] | bd);
}

static void tcg_out_bc_lab(TCGContext *s, TCGCond cond, TCGLabel *l)
{
    int bd = 0;
    if (l->has_value) {
        bd = reloc_pc14_val(tcg_splitwx_to_rx(s->code_ptr), l->u.value_ptr);
    } else {
        tcg_out_reloc(s, s->code_ptr, R_PPC_REL14, l, 0);
    }
    tcg_out_bc(s, cond, bd);
}

static void tgen_brcond(TCGContext *s, TCGType type, TCGCond cond,
                        TCGReg arg1, TCGReg arg2, TCGLabel *l)
{
    tcg_out_cmp(s, cond, arg1, arg2, false, 0, type);
    tcg_out_bc_lab(s, cond, l);
}

static void tgen_brcondi(TCGContext *s, TCGType type, TCGCond cond,
                         TCGReg arg1, tcg_target_long arg2, TCGLabel *l)
{
    tcg_out_cmp(s, cond, arg1, arg2, true, 0, type);
    tcg_out_bc_lab(s, cond, l);
}

static const TCGOutOpBrcond outop_brcond = {
    .base.static_constraint = C_O0_I2(r, rC),
    .out_rr = tgen_brcond,
    .out_ri = tgen_brcondi,
};

static void tgen_movcond(TCGContext *s, TCGType type, TCGCond cond,
                         TCGReg dest, TCGReg c1, TCGArg c2, bool const_c2,
                         TCGArg v1, bool const_v1, TCGArg v2, bool const_v2)
{
    /* If for some reason both inputs are zero, don't produce bad code.  */
    if (v1 == 0 && v2 == 0) {
        tcg_out_movi(s, type, dest, 0);
        return;
    }

    tcg_out_cmp(s, cond, c1, c2, const_c2, 0, type);

    if (have_isel) {
        int isel = tcg_to_isel[cond];

        /* Swap the V operands if the operation indicates inversion.  */
        if (isel & 1) {
            int t = v1;
            v1 = v2;
            v2 = t;
            isel &= ~1;
        }
        /* V1 == 0 is handled by isel; V2 == 0 must be handled by hand.  */
        if (v2 == 0) {
            tcg_out_movi(s, type, TCG_REG_R0, 0);
        }
        tcg_out32(s, isel | TAB(dest, v1, v2));
    } else {
        if (dest == v2) {
            cond = tcg_invert_cond(cond);
            v2 = v1;
        } else if (dest != v1) {
            if (v1 == 0) {
                tcg_out_movi(s, type, dest, 0);
            } else {
                tcg_out_mov(s, type, dest, v1);
            }
        }
        /* Branch forward over one insn */
        tcg_out_bc(s, cond, 8);
        if (v2 == 0) {
            tcg_out_movi(s, type, dest, 0);
        } else {
            tcg_out_mov(s, type, dest, v2);
        }
    }
}

static const TCGOutOpMovcond outop_movcond = {
    .base.static_constraint = C_O1_I4(r, r, rC, rZ, rZ),
    .out = tgen_movcond,
};

static void tcg_out_cntxz(TCGContext *s, TCGType type, uint32_t opc,
                          TCGArg a0, TCGArg a1, TCGArg a2, bool const_a2)
{
    if (const_a2 && a2 == (type == TCG_TYPE_I32 ? 32 : 64)) {
        tcg_out32(s, opc | RA(a0) | RS(a1));
    } else {
        tcg_out_cmp(s, TCG_COND_EQ, a1, 0, 1, 0, type);
        /* Note that the only other valid constant for a2 is 0.  */
        if (have_isel) {
            tcg_out32(s, opc | RA(TCG_REG_R0) | RS(a1));
            tcg_out32(s, tcg_to_isel[TCG_COND_EQ] | TAB(a0, a2, TCG_REG_R0));
        } else if (!const_a2 && a0 == a2) {
            tcg_out_bc(s, TCG_COND_EQ, 8);
            tcg_out32(s, opc | RA(a0) | RS(a1));
        } else {
            tcg_out32(s, opc | RA(a0) | RS(a1));
            tcg_out_bc(s, TCG_COND_NE, 8);
            if (const_a2) {
                tcg_out_movi(s, type, a0, 0);
            } else {
                tcg_out_mov(s, type, a0, a2);
            }
        }
    }
}

static void tcg_out_cmp2(TCGContext *s, TCGCond cond, TCGReg al, TCGReg ah,
                         TCGArg bl, bool blconst, TCGArg bh, bool bhconst)
{
    static const struct { uint8_t bit1, bit2; } bits[] = {
        [TCG_COND_LT ] = { CR_LT, CR_LT },
        [TCG_COND_LE ] = { CR_LT, CR_GT },
        [TCG_COND_GT ] = { CR_GT, CR_GT },
        [TCG_COND_GE ] = { CR_GT, CR_LT },
        [TCG_COND_LTU] = { CR_LT, CR_LT },
        [TCG_COND_LEU] = { CR_LT, CR_GT },
        [TCG_COND_GTU] = { CR_GT, CR_GT },
        [TCG_COND_GEU] = { CR_GT, CR_LT },
    };

    TCGCond cond2;
    int op, bit1, bit2;

    switch (cond) {
    case TCG_COND_EQ:
        op = CRAND;
        goto do_equality;
    case TCG_COND_NE:
        op = CRNAND;
    do_equality:
        tcg_out_cmp(s, cond, al, bl, blconst, 6, TCG_TYPE_I32);
        tcg_out_cmp(s, cond, ah, bh, bhconst, 7, TCG_TYPE_I32);
        tcg_out32(s, op | BT(0, CR_EQ) | BA(6, CR_EQ) | BB(7, CR_EQ));
        break;

    case TCG_COND_TSTEQ:
    case TCG_COND_TSTNE:
        if (blconst) {
            tcg_out_andi32(s, TCG_REG_R0, al, bl);
        } else {
            tcg_out32(s, AND | SAB(al, TCG_REG_R0, bl));
        }
        if (bhconst) {
            tcg_out_andi32(s, TCG_REG_TMP1, ah, bh);
        } else {
            tcg_out32(s, AND | SAB(ah, TCG_REG_TMP1, bh));
        }
        tcg_out32(s, OR | SAB(TCG_REG_R0, TCG_REG_R0, TCG_REG_TMP1) | 1);
        break;

    case TCG_COND_LT:
    case TCG_COND_LE:
    case TCG_COND_GT:
    case TCG_COND_GE:
    case TCG_COND_LTU:
    case TCG_COND_LEU:
    case TCG_COND_GTU:
    case TCG_COND_GEU:
        bit1 = bits[cond].bit1;
        bit2 = bits[cond].bit2;
        op = (bit1 != bit2 ? CRANDC : CRAND);
        cond2 = tcg_unsigned_cond(cond);

        tcg_out_cmp(s, cond, ah, bh, bhconst, 6, TCG_TYPE_I32);
        tcg_out_cmp(s, cond2, al, bl, blconst, 7, TCG_TYPE_I32);
        tcg_out32(s, op | BT(0, CR_EQ) | BA(6, CR_EQ) | BB(7, bit2));
        tcg_out32(s, CROR | BT(0, CR_EQ) | BA(6, bit1) | BB(0, CR_EQ));
        break;

    default:
        g_assert_not_reached();
    }
}

static void tgen_setcond2(TCGContext *s, TCGCond cond, TCGReg ret,
                          TCGReg al, TCGReg ah,
                          TCGArg bl, bool const_bl,
                          TCGArg bh, bool const_bh)
{
    tcg_out_cmp2(s, cond, al, ah, bl, const_bl, bh, const_bh);
    tcg_out32(s, MFOCRF | RT(TCG_REG_R0) | FXM(0));
    tcg_out_rlw(s, RLWINM, ret, TCG_REG_R0, CR_EQ + 0*4 + 1, 31, 31);
}

#if TCG_TARGET_REG_BITS != 32
__attribute__((unused))
#endif
static const TCGOutOpSetcond2 outop_setcond2 = {
    .base.static_constraint = C_O1_I4(r, r, r, rU, rC),
    .out = tgen_setcond2,
};

static void tgen_brcond2(TCGContext *s, TCGCond cond, TCGReg al, TCGReg ah,
                         TCGArg bl, bool const_bl,
                         TCGArg bh, bool const_bh, TCGLabel *l)
{
    assert(TCG_TARGET_REG_BITS == 32);
    tcg_out_cmp2(s, cond, al, ah, bl, const_bl, bh, const_bh);
    tcg_out_bc_lab(s, TCG_COND_EQ, l);
}

#if TCG_TARGET_REG_BITS != 32
__attribute__((unused))
#endif
static const TCGOutOpBrcond2 outop_brcond2 = {
    .base.static_constraint = C_O0_I4(r, r, rU, rC),
    .out = tgen_brcond2,
};

static void tcg_out_mb(TCGContext *s, TCGArg a0)
{
    uint32_t insn;

    if (a0 & TCG_MO_ST_LD) {
        insn = HWSYNC;
    } else {
        insn = LWSYNC;
    }

    tcg_out32(s, insn);
}

static void tcg_out_call_int(TCGContext *s, int lk,
                             const tcg_insn_unit *target)
{
#ifdef _CALL_AIX
    /* Look through the descriptor.  If the branch is in range, and we
       don't have to spend too much effort on building the toc.  */
    const void *tgt = ((const void * const *)target)[0];
    uintptr_t toc = ((const uintptr_t *)target)[1];
    intptr_t diff = tcg_pcrel_diff(s, tgt);

    if (in_range_b(diff) && toc == (uint32_t)toc) {
        tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP1, toc);
        tcg_out_b(s, lk, tgt);
    } else {
        /* Fold the low bits of the constant into the addresses below.  */
        intptr_t arg = (intptr_t)target;
        int ofs = (int16_t)arg;

        if (ofs + 8 < 0x8000) {
            arg -= ofs;
        } else {
            ofs = 0;
        }
        tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP1, arg);
        tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_R0, TCG_REG_TMP1, ofs);
        tcg_out32(s, MTSPR | RA(TCG_REG_R0) | CTR);
        tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_R2, TCG_REG_TMP1, ofs + SZP);
        tcg_out32(s, BCCTR | BO_ALWAYS | lk);
    }
#elif defined(_CALL_ELF) && _CALL_ELF == 2
    intptr_t diff;

    /* In the ELFv2 ABI, we have to set up r12 to contain the destination
       address, which the callee uses to compute its TOC address.  */
    /* FIXME: when the branch is in range, we could avoid r12 load if we
       knew that the destination uses the same TOC, and what its local
       entry point offset is.  */
    tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R12, (intptr_t)target);

    diff = tcg_pcrel_diff(s, target);
    if (in_range_b(diff)) {
        tcg_out_b(s, lk, target);
    } else {
        tcg_out32(s, MTSPR | RS(TCG_REG_R12) | CTR);
        tcg_out32(s, BCCTR | BO_ALWAYS | lk);
    }
#else
    tcg_out_b(s, lk, target);
#endif
}

static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target,
                         const TCGHelperInfo *info)
{
    tcg_out_call_int(s, LK, target);
}

static const uint32_t qemu_ldx_opc[(MO_SSIZE + MO_BSWAP) + 1] = {
    [MO_UB] = LBZX,
    [MO_UW] = LHZX,
    [MO_UL] = LWZX,
    [MO_UQ] = LDX,
    [MO_SW] = LHAX,
    [MO_SL] = LWAX,
    [MO_BSWAP | MO_UB] = LBZX,
    [MO_BSWAP | MO_UW] = LHBRX,
    [MO_BSWAP | MO_UL] = LWBRX,
    [MO_BSWAP | MO_UQ] = LDBRX,
};

static const uint32_t qemu_stx_opc[(MO_SIZE + MO_BSWAP) + 1] = {
    [MO_UB] = STBX,
    [MO_UW] = STHX,
    [MO_UL] = STWX,
    [MO_UQ] = STDX,
    [MO_BSWAP | MO_UB] = STBX,
    [MO_BSWAP | MO_UW] = STHBRX,
    [MO_BSWAP | MO_UL] = STWBRX,
    [MO_BSWAP | MO_UQ] = STDBRX,
};

static TCGReg ldst_ra_gen(TCGContext *s, const TCGLabelQemuLdst *l, int arg)
{
    if (arg < 0) {
        arg = TCG_REG_TMP1;
    }
    tcg_out32(s, MFSPR | RT(arg) | LR);
    return arg;
}

/*
 * For the purposes of ppc32 sorting 4 input registers into 4 argument
 * registers, there is an outside chance we would require 3 temps.
 */
static const TCGLdstHelperParam ldst_helper_param = {
    .ra_gen = ldst_ra_gen,
    .ntmp = 3,
    .tmp = { TCG_REG_TMP1, TCG_REG_TMP2, TCG_REG_R0 }
};

static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
{
    MemOp opc = get_memop(lb->oi);

    if (!reloc_pc14(lb->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) {
        return false;
    }

    tcg_out_ld_helper_args(s, lb, &ldst_helper_param);
    tcg_out_call_int(s, LK, qemu_ld_helpers[opc & MO_SIZE]);
    tcg_out_ld_helper_ret(s, lb, false, &ldst_helper_param);

    tcg_out_b(s, 0, lb->raddr);
    return true;
}

static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
{
    MemOp opc = get_memop(lb->oi);

    if (!reloc_pc14(lb->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) {
        return false;
    }

    tcg_out_st_helper_args(s, lb, &ldst_helper_param);
    tcg_out_call_int(s, LK, qemu_st_helpers[opc & MO_SIZE]);

    tcg_out_b(s, 0, lb->raddr);
    return true;
}

typedef struct {
    TCGReg base;
    TCGReg index;
    TCGAtomAlign aa;
} HostAddress;

bool tcg_target_has_memory_bswap(MemOp memop)
{
    TCGAtomAlign aa;

    if ((memop & MO_SIZE) <= MO_64) {
        return true;
    }

    /*
     * Reject 16-byte memop with 16-byte atomicity,
     * but do allow a pair of 64-bit operations.
     */
    aa = atom_and_align_for_opc(tcg_ctx, memop, MO_ATOM_IFALIGN, true);
    return aa.atom <= MO_64;
}

/* We expect to use a 16-bit negative offset from ENV.  */
#define MIN_TLB_MASK_TABLE_OFS  -32768

/*
 * For system-mode, perform the TLB load and compare.
 * For user-mode, perform any required alignment tests.
 * In both cases, return a TCGLabelQemuLdst structure if the slow path
 * is required and fill in @h with the host address for the fast path.
 */
static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h,
                                           TCGReg addr, MemOpIdx oi, bool is_ld)
{
    TCGType addr_type = s->addr_type;
    TCGLabelQemuLdst *ldst = NULL;
    MemOp opc = get_memop(oi);
    MemOp a_bits, s_bits;

    /*
     * Book II, Section 1.4, Single-Copy Atomicity, specifies:
     *
     * Before 3.0, "An access that is not atomic is performed as a set of
     * smaller disjoint atomic accesses. In general, the number and alignment
     * of these accesses are implementation-dependent."  Thus MO_ATOM_IFALIGN.
     *
     * As of 3.0, "the non-atomic access is performed as described in
     * the corresponding list", which matches MO_ATOM_SUBALIGN.
     */
    s_bits = opc & MO_SIZE;
    h->aa = atom_and_align_for_opc(s, opc,
                                   have_isa_3_00 ? MO_ATOM_SUBALIGN
                                                 : MO_ATOM_IFALIGN,
                                   s_bits == MO_128);
    a_bits = h->aa.align;

    if (tcg_use_softmmu) {
        int mem_index = get_mmuidx(oi);
        int cmp_off = is_ld ? offsetof(CPUTLBEntry, addr_read)
                            : offsetof(CPUTLBEntry, addr_write);
        int fast_off = tlb_mask_table_ofs(s, mem_index);
        int mask_off = fast_off + offsetof(CPUTLBDescFast, mask);
        int table_off = fast_off + offsetof(CPUTLBDescFast, table);

        ldst = new_ldst_label(s);
        ldst->is_ld = is_ld;
        ldst->oi = oi;
        ldst->addr_reg = addr;

        /* Load tlb_mask[mmu_idx] and tlb_table[mmu_idx].  */
        tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP1, TCG_AREG0, mask_off);
        tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP2, TCG_AREG0, table_off);

        /* Extract the page index, shifted into place for tlb index.  */
        if (TCG_TARGET_REG_BITS == 32) {
            tcg_out_shri32(s, TCG_REG_R0, addr,
                           s->page_bits - CPU_TLB_ENTRY_BITS);
        } else {
            tcg_out_shri64(s, TCG_REG_R0, addr,
                           s->page_bits - CPU_TLB_ENTRY_BITS);
        }
        tcg_out32(s, AND | SAB(TCG_REG_TMP1, TCG_REG_TMP1, TCG_REG_R0));

        /*
         * Load the TLB comparator into TMP2.
         * For 64-bit host, always load the entire 64-bit slot for simplicity.
         * We will ignore the high bits with tcg_out_cmp(..., addr_type).
         */
        if (cmp_off == 0) {
            tcg_out32(s, (TCG_TARGET_REG_BITS == 64 ? LDUX : LWZUX)
                      | TAB(TCG_REG_TMP2, TCG_REG_TMP1, TCG_REG_TMP2));
        } else {
            tcg_out32(s, ADD | TAB(TCG_REG_TMP1, TCG_REG_TMP1, TCG_REG_TMP2));
            tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP2, TCG_REG_TMP1, cmp_off);
        }

        /*
         * Load the TLB addend for use on the fast path.
         * Do this asap to minimize any load use delay.
         */
        if (TCG_TARGET_REG_BITS == 64 || addr_type == TCG_TYPE_I32) {
            tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP1, TCG_REG_TMP1,
                       offsetof(CPUTLBEntry, addend));
        }

        /* Clear the non-page, non-alignment bits from the address in R0. */
        if (TCG_TARGET_REG_BITS == 32) {
            /*
             * We don't support unaligned accesses on 32-bits.
             * Preserve the bottom bits and thus trigger a comparison
             * failure on unaligned accesses.
             */
            if (a_bits < s_bits) {
                a_bits = s_bits;
            }
            tcg_out_rlw(s, RLWINM, TCG_REG_R0, addr, 0,
                        (32 - a_bits) & 31, 31 - s->page_bits);
        } else {
            TCGReg t = addr;

            /*
             * If the access is unaligned, we need to make sure we fail if we
             * cross a page boundary.  The trick is to add the access size-1
             * to the address before masking the low bits.  That will make the
             * address overflow to the next page if we cross a page boundary,
             * which will then force a mismatch of the TLB compare.
             */
            if (a_bits < s_bits) {
                unsigned a_mask = (1 << a_bits) - 1;
                unsigned s_mask = (1 << s_bits) - 1;
                tcg_out32(s, ADDI | TAI(TCG_REG_R0, t, s_mask - a_mask));
                t = TCG_REG_R0;
            }

            /* Mask the address for the requested alignment.  */
            if (addr_type == TCG_TYPE_I32) {
                tcg_out_rlw(s, RLWINM, TCG_REG_R0, t, 0,
                            (32 - a_bits) & 31, 31 - s->page_bits);
            } else if (a_bits == 0) {
                tcg_out_rld(s, RLDICR, TCG_REG_R0, t, 0, 63 - s->page_bits);
            } else {
                tcg_out_rld(s, RLDICL, TCG_REG_R0, t,
                            64 - s->page_bits, s->page_bits - a_bits);
                tcg_out_rld(s, RLDICL, TCG_REG_R0, TCG_REG_R0, s->page_bits, 0);
            }
        }

        /* Full comparison into cr0. */
        tcg_out_cmp(s, TCG_COND_EQ, TCG_REG_R0, TCG_REG_TMP2, 0, 0, addr_type);

        /* Load a pointer into the current opcode w/conditional branch-link. */
        ldst->label_ptr[0] = s->code_ptr;
        tcg_out_bc(s, TCG_COND_NE, LK);

        h->base = TCG_REG_TMP1;
    } else {
        if (a_bits) {
            ldst = new_ldst_label(s);
            ldst->is_ld = is_ld;
            ldst->oi = oi;
            ldst->addr_reg = addr;

            /* We are expecting a_bits to max out at 7, much lower than ANDI. */
            tcg_debug_assert(a_bits < 16);
            tcg_out32(s, ANDI | SAI(addr, TCG_REG_R0, (1 << a_bits) - 1));

            ldst->label_ptr[0] = s->code_ptr;
            tcg_out32(s, BC | BI(0, CR_EQ) | BO_COND_FALSE | LK);
        }

        h->base = guest_base ? TCG_GUEST_BASE_REG : 0;
    }

    if (TCG_TARGET_REG_BITS == 64 && addr_type == TCG_TYPE_I32) {
        /* Zero-extend the guest address for use in the host address. */
        tcg_out_ext32u(s, TCG_REG_TMP2, addr);
        h->index = TCG_REG_TMP2;
    } else {
        h->index = addr;
    }

    return ldst;
}

static void tcg_out_qemu_ld(TCGContext *s, TCGReg datalo, TCGReg datahi,
                            TCGReg addr, MemOpIdx oi, TCGType data_type)
{
    MemOp opc = get_memop(oi);
    TCGLabelQemuLdst *ldst;
    HostAddress h;

    ldst = prepare_host_addr(s, &h, addr, oi, true);

    if (TCG_TARGET_REG_BITS == 32 && (opc & MO_SIZE) == MO_64) {
        if (opc & MO_BSWAP) {
            tcg_out32(s, ADDI | TAI(TCG_REG_R0, h.index, 4));
            tcg_out32(s, LWBRX | TAB(datalo, h.base, h.index));
            tcg_out32(s, LWBRX | TAB(datahi, h.base, TCG_REG_R0));
        } else if (h.base != 0) {
            tcg_out32(s, ADDI | TAI(TCG_REG_R0, h.index, 4));
            tcg_out32(s, LWZX | TAB(datahi, h.base, h.index));
            tcg_out32(s, LWZX | TAB(datalo, h.base, TCG_REG_R0));
        } else if (h.index == datahi) {
            tcg_out32(s, LWZ | TAI(datalo, h.index, 4));
            tcg_out32(s, LWZ | TAI(datahi, h.index, 0));
        } else {
            tcg_out32(s, LWZ | TAI(datahi, h.index, 0));
            tcg_out32(s, LWZ | TAI(datalo, h.index, 4));
        }
    } else {
        uint32_t insn = qemu_ldx_opc[opc & (MO_BSWAP | MO_SSIZE)];
        if (!have_isa_2_06 && insn == LDBRX) {
            tcg_out32(s, ADDI | TAI(TCG_REG_R0, h.index, 4));
            tcg_out32(s, LWBRX | TAB(datalo, h.base, h.index));
            tcg_out32(s, LWBRX | TAB(TCG_REG_R0, h.base, TCG_REG_R0));
            tcg_out_rld(s, RLDIMI, datalo, TCG_REG_R0, 32, 0);
        } else if (insn) {
            tcg_out32(s, insn | TAB(datalo, h.base, h.index));
        } else {
            insn = qemu_ldx_opc[opc & (MO_SIZE | MO_BSWAP)];
            tcg_out32(s, insn | TAB(datalo, h.base, h.index));
            tcg_out_movext(s, TCG_TYPE_REG, datalo,
                           TCG_TYPE_REG, opc & MO_SSIZE, datalo);
        }
    }

    if (ldst) {
        ldst->type = data_type;
        ldst->datalo_reg = datalo;
        ldst->datahi_reg = datahi;
        ldst->raddr = tcg_splitwx_to_rx(s->code_ptr);
    }
}

static void tcg_out_qemu_st(TCGContext *s, TCGReg datalo, TCGReg datahi,
                            TCGReg addr, MemOpIdx oi, TCGType data_type)
{
    MemOp opc = get_memop(oi);
    TCGLabelQemuLdst *ldst;
    HostAddress h;

    ldst = prepare_host_addr(s, &h, addr, oi, false);

    if (TCG_TARGET_REG_BITS == 32 && (opc & MO_SIZE) == MO_64) {
        if (opc & MO_BSWAP) {
            tcg_out32(s, ADDI | TAI(TCG_REG_R0, h.index, 4));
            tcg_out32(s, STWBRX | SAB(datalo, h.base, h.index));
            tcg_out32(s, STWBRX | SAB(datahi, h.base, TCG_REG_R0));
        } else if (h.base != 0) {
            tcg_out32(s, ADDI | TAI(TCG_REG_R0, h.index, 4));
            tcg_out32(s, STWX | SAB(datahi, h.base, h.index));
            tcg_out32(s, STWX | SAB(datalo, h.base, TCG_REG_R0));
        } else {
            tcg_out32(s, STW | TAI(datahi, h.index, 0));
            tcg_out32(s, STW | TAI(datalo, h.index, 4));
        }
    } else {
        uint32_t insn = qemu_stx_opc[opc & (MO_BSWAP | MO_SIZE)];
        if (!have_isa_2_06 && insn == STDBRX) {
            tcg_out32(s, STWBRX | SAB(datalo, h.base, h.index));
            tcg_out32(s, ADDI | TAI(TCG_REG_TMP2, h.index, 4));
            tcg_out_shri64(s, TCG_REG_R0, datalo, 32);
            tcg_out32(s, STWBRX | SAB(TCG_REG_R0, h.base, TCG_REG_TMP2));
        } else {
            tcg_out32(s, insn | SAB(datalo, h.base, h.index));
        }
    }

    if (ldst) {
        ldst->type = data_type;
        ldst->datalo_reg = datalo;
        ldst->datahi_reg = datahi;
        ldst->raddr = tcg_splitwx_to_rx(s->code_ptr);
    }
}

static void tcg_out_qemu_ldst_i128(TCGContext *s, TCGReg datalo, TCGReg datahi,
                                   TCGReg addr_reg, MemOpIdx oi, bool is_ld)
{
    TCGLabelQemuLdst *ldst;
    HostAddress h;
    bool need_bswap;
    uint32_t insn;
    TCGReg index;

    ldst = prepare_host_addr(s, &h, addr_reg, oi, is_ld);

    /* Compose the final address, as LQ/STQ have no indexing. */
    index = h.index;
    if (h.base != 0) {
        index = TCG_REG_TMP1;
        tcg_out32(s, ADD | TAB(index, h.base, h.index));
    }
    need_bswap = get_memop(oi) & MO_BSWAP;

    if (h.aa.atom == MO_128) {
        tcg_debug_assert(!need_bswap);
        tcg_debug_assert(datalo & 1);
        tcg_debug_assert(datahi == datalo - 1);
        tcg_debug_assert(!is_ld || datahi != index);
        insn = is_ld ? LQ : STQ;
        tcg_out32(s, insn | TAI(datahi, index, 0));
    } else {
        TCGReg d1, d2;

        if (HOST_BIG_ENDIAN ^ need_bswap) {
            d1 = datahi, d2 = datalo;
        } else {
            d1 = datalo, d2 = datahi;
        }

        if (need_bswap) {
            tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R0, 8);
            insn = is_ld ? LDBRX : STDBRX;
            tcg_out32(s, insn | TAB(d1, 0, index));
            tcg_out32(s, insn | TAB(d2, index, TCG_REG_R0));
        } else {
            insn = is_ld ? LD : STD;
            tcg_out32(s, insn | TAI(d1, index, 0));
            tcg_out32(s, insn | TAI(d2, index, 8));
        }
    }

    if (ldst) {
        ldst->type = TCG_TYPE_I128;
        ldst->datalo_reg = datalo;
        ldst->datahi_reg = datahi;
        ldst->raddr = tcg_splitwx_to_rx(s->code_ptr);
    }
}

static void tcg_out_nop_fill(tcg_insn_unit *p, int count)
{
    int i;
    for (i = 0; i < count; ++i) {
        p[i] = NOP;
    }
}

/* Parameters for function call generation, used in tcg.c.  */
#define TCG_TARGET_STACK_ALIGN       16

#ifdef _CALL_AIX
# define LINK_AREA_SIZE                (6 * SZR)
# define LR_OFFSET                     (1 * SZR)
# define TCG_TARGET_CALL_STACK_OFFSET  (LINK_AREA_SIZE + 8 * SZR)
#elif defined(_CALL_DARWIN)
# define LINK_AREA_SIZE                (6 * SZR)
# define LR_OFFSET                     (2 * SZR)
#elif TCG_TARGET_REG_BITS == 64
# if defined(_CALL_ELF) && _CALL_ELF == 2
#  define LINK_AREA_SIZE               (4 * SZR)
#  define LR_OFFSET                    (1 * SZR)
# endif
#else /* TCG_TARGET_REG_BITS == 32 */
# if defined(_CALL_SYSV)
#  define LINK_AREA_SIZE               (2 * SZR)
#  define LR_OFFSET                    (1 * SZR)
# endif
#endif
#ifndef LR_OFFSET
# error "Unhandled abi"
#endif
#ifndef TCG_TARGET_CALL_STACK_OFFSET
# define TCG_TARGET_CALL_STACK_OFFSET  LINK_AREA_SIZE
#endif

#define CPU_TEMP_BUF_SIZE  (CPU_TEMP_BUF_NLONGS * (int)sizeof(long))
#define REG_SAVE_SIZE      ((int)ARRAY_SIZE(tcg_target_callee_save_regs) * SZR)

#define FRAME_SIZE ((TCG_TARGET_CALL_STACK_OFFSET   \
                     + TCG_STATIC_CALL_ARGS_SIZE    \
                     + CPU_TEMP_BUF_SIZE            \
                     + REG_SAVE_SIZE                \
                     + TCG_TARGET_STACK_ALIGN - 1)  \
                    & -TCG_TARGET_STACK_ALIGN)

#define REG_SAVE_BOT (FRAME_SIZE - REG_SAVE_SIZE)

static void tcg_target_qemu_prologue(TCGContext *s)
{
    int i;

#ifdef _CALL_AIX
    const void **desc = (const void **)s->code_ptr;
    desc[0] = tcg_splitwx_to_rx(desc + 2);  /* entry point */
    desc[1] = 0;                            /* environment pointer */
    s->code_ptr = (void *)(desc + 2);       /* skip over descriptor */
#endif

    tcg_set_frame(s, TCG_REG_CALL_STACK, REG_SAVE_BOT - CPU_TEMP_BUF_SIZE,
                  CPU_TEMP_BUF_SIZE);

    /* Prologue */
    tcg_out32(s, MFSPR | RT(TCG_REG_R0) | LR);
    tcg_out32(s, (SZR == 8 ? STDU : STWU)
              | SAI(TCG_REG_R1, TCG_REG_R1, -FRAME_SIZE));

    for (i = 0; i < ARRAY_SIZE(tcg_target_callee_save_regs); ++i) {
        tcg_out_st(s, TCG_TYPE_REG, tcg_target_callee_save_regs[i],
                   TCG_REG_R1, REG_SAVE_BOT + i * SZR);
    }
    tcg_out_st(s, TCG_TYPE_PTR, TCG_REG_R0, TCG_REG_R1, FRAME_SIZE+LR_OFFSET);

    if (!tcg_use_softmmu && guest_base) {
        tcg_out_movi_int(s, TCG_TYPE_PTR, TCG_GUEST_BASE_REG, guest_base, true);
        tcg_regset_set_reg(s->reserved_regs, TCG_GUEST_BASE_REG);
    }

    tcg_out_mov(s, TCG_TYPE_PTR, TCG_AREG0, tcg_target_call_iarg_regs[0]);
    tcg_out32(s, MTSPR | RS(tcg_target_call_iarg_regs[1]) | CTR);
    tcg_out32(s, BCCTR | BO_ALWAYS);

    /* Epilogue */
    tcg_code_gen_epilogue = tcg_splitwx_to_rx(s->code_ptr);

    tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_R0, TCG_REG_R1, FRAME_SIZE+LR_OFFSET);
    for (i = 0; i < ARRAY_SIZE(tcg_target_callee_save_regs); ++i) {
        tcg_out_ld(s, TCG_TYPE_REG, tcg_target_callee_save_regs[i],
                   TCG_REG_R1, REG_SAVE_BOT + i * SZR);
    }
    tcg_out32(s, MTSPR | RS(TCG_REG_R0) | LR);
    tcg_out32(s, ADDI | TAI(TCG_REG_R1, TCG_REG_R1, FRAME_SIZE));
    tcg_out32(s, BCLR | BO_ALWAYS);
}

static void tcg_out_tb_start(TCGContext *s)
{
    /* Load TCG_REG_TB. */
    if (USE_REG_TB) {
        if (have_isa_3_00) {
            /* lnia REG_TB */
            tcg_out_addpcis(s, TCG_REG_TB, 0);
        } else {
            /* bcl 20,31,$+4 (preferred form for getting nia) */
            tcg_out32(s, BC | BO_ALWAYS | BI(7, CR_SO) | 0x4 | LK);
            tcg_out32(s, MFSPR | RT(TCG_REG_TB) | LR);
        }
    }
}

static void tcg_out_exit_tb(TCGContext *s, uintptr_t arg)
{
    tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R3, arg);
    tcg_out_b(s, 0, tcg_code_gen_epilogue);
}

static void tcg_out_goto_tb(TCGContext *s, int which)
{
    uintptr_t ptr = get_jmp_target_addr(s, which);
    int16_t lo;

    /* Direct branch will be patched by tb_target_set_jmp_target. */
    set_jmp_insn_offset(s, which);
    tcg_out32(s, NOP);

    /* When branch is out of range, fall through to indirect. */
    if (USE_REG_TB) {
        ptrdiff_t offset = ppc_tbrel_diff(s, (void *)ptr);
        tcg_out_mem_long(s, LD, LDX, TCG_REG_TMP1, TCG_REG_TB, offset);
    } else if (have_isa_3_10) {
        ptrdiff_t offset = tcg_pcrel_diff_for_prefix(s, (void *)ptr);
        tcg_out_8ls_d(s, PLD, TCG_REG_TMP1, 0, offset, 1);
    } else if (have_isa_3_00) {
        ptrdiff_t offset = tcg_pcrel_diff(s, (void *)ptr) - 4;
        lo = offset;
        tcg_out_addpcis(s, TCG_REG_TMP1, offset - lo);
        tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP1, TCG_REG_TMP1, lo);
    } else {
        lo = ptr;
        tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP1, ptr - lo);
        tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP1, TCG_REG_TMP1, lo);
    }

    tcg_out32(s, MTSPR | RS(TCG_REG_TMP1) | CTR);
    tcg_out32(s, BCCTR | BO_ALWAYS);
    set_jmp_reset_offset(s, which);
}

void tb_target_set_jmp_target(const TranslationBlock *tb, int n,
                              uintptr_t jmp_rx, uintptr_t jmp_rw)
{
    uintptr_t addr = tb->jmp_target_addr[n];
    intptr_t diff = addr - jmp_rx;
    tcg_insn_unit insn;

    if (in_range_b(diff)) {
        insn = B | (diff & 0x3fffffc);
    } else {
        insn = NOP;
    }

    qatomic_set((uint32_t *)jmp_rw, insn);
    flush_idcache_range(jmp_rx, jmp_rw, 4);
}


static void tgen_add(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, ADD | TAB(a0, a1, a2));
}

static void tgen_addi(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    tcg_out_mem_long(s, ADDI, ADD, a0, a1, a2);
}

static const TCGOutOpBinary outop_add = {
    .base.static_constraint = C_O1_I2(r, r, rT),
    .out_rrr = tgen_add,
    .out_rri = tgen_addi,
};

static void tgen_and(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, AND | SAB(a1, a0, a2));
}

static void tgen_andi(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    if (type == TCG_TYPE_I32) {
        tcg_out_andi32(s, a0, a1, a2);
    } else {
        tcg_out_andi64(s, a0, a1, a2);
    }
}

static const TCGOutOpBinary outop_and = {
    .base.static_constraint = C_O1_I2(r, r, ri),
    .out_rrr = tgen_and,
    .out_rri = tgen_andi,
};

static void tgen_andc(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, ANDC | SAB(a1, a0, a2));
}

static const TCGOutOpBinary outop_andc = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_andc,
};

static void tgen_clz(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? CNTLZW : CNTLZD;
    tcg_out_cntxz(s, type, insn, a0, a1, a2, false);
}

static void tgen_clzi(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? CNTLZW : CNTLZD;
    tcg_out_cntxz(s, type, insn, a0, a1, a2, true);
}

static const TCGOutOpBinary outop_clz = {
    .base.static_constraint = C_O1_I2(r, r, rZW),
    .out_rrr = tgen_clz,
    .out_rri = tgen_clzi,
};

static void tgen_ctpop(TCGContext *s, TCGType type, TCGReg a0, TCGReg a1)
{
    uint32_t insn = type == TCG_TYPE_I32 ? CNTPOPW : CNTPOPD;
    tcg_out32(s, insn | SAB(a1, a0, 0));
}

static TCGConstraintSetIndex cset_ctpop(TCGType type, unsigned flags)
{
    return have_isa_2_06 ? C_O1_I1(r, r) : C_NotImplemented;
}

static const TCGOutOpUnary outop_ctpop = {
    .base.static_constraint = C_Dynamic,
    .base.dynamic_constraint = cset_ctpop,
    .out_rr = tgen_ctpop,
};

static void tgen_ctz(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? CNTTZW : CNTTZD;
    tcg_out_cntxz(s, type, insn, a0, a1, a2, false);
}

static void tgen_ctzi(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? CNTTZW : CNTTZD;
    tcg_out_cntxz(s, type, insn, a0, a1, a2, true);
}

static TCGConstraintSetIndex cset_ctz(TCGType type, unsigned flags)
{
    return have_isa_3_00 ? C_O1_I2(r, r, rZW) : C_NotImplemented;
}

static const TCGOutOpBinary outop_ctz = {
    .base.static_constraint = C_Dynamic,
    .base.dynamic_constraint = cset_ctz,
    .out_rrr = tgen_ctz,
    .out_rri = tgen_ctzi,
};

static void tgen_eqv(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, EQV | SAB(a1, a0, a2));
}

static void tgen_divs(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? DIVW : DIVD;
    tcg_out32(s, insn | TAB(a0, a1, a2));
}

static const TCGOutOpBinary outop_divs = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_divs,
};

static const TCGOutOpDivRem outop_divs2 = {
    .base.static_constraint = C_NotImplemented,
};

static void tgen_divu(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? DIVWU : DIVDU;
    tcg_out32(s, insn | TAB(a0, a1, a2));
}

static const TCGOutOpBinary outop_divu = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_divu,
};

static const TCGOutOpDivRem outop_divu2 = {
    .base.static_constraint = C_NotImplemented,
};

static const TCGOutOpBinary outop_eqv = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_eqv,
};

static void tgen_mul(TCGContext *s, TCGType type,
                    TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? MULLW : MULLD;
    tcg_out32(s, insn | TAB(a0, a1, a2));
}

static void tgen_muli(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    tcg_out32(s, MULLI | TAI(a0, a1, a2));
}

static const TCGOutOpBinary outop_mul = {
    .base.static_constraint = C_O1_I2(r, r, rI),
    .out_rrr = tgen_mul,
    .out_rri = tgen_muli,
};

static const TCGOutOpMul2 outop_muls2 = {
    .base.static_constraint = C_NotImplemented,
};

static void tgen_mulsh(TCGContext *s, TCGType type,
                       TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? MULHW : MULHD;
    tcg_out32(s, insn | TAB(a0, a1, a2));
}

static const TCGOutOpBinary outop_mulsh = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_mulsh,
};

static const TCGOutOpMul2 outop_mulu2 = {
    .base.static_constraint = C_NotImplemented,
};

static void tgen_muluh(TCGContext *s, TCGType type,
                       TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? MULHWU : MULHDU;
    tcg_out32(s, insn | TAB(a0, a1, a2));
}

static const TCGOutOpBinary outop_muluh = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_muluh,
};

static void tgen_nand(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, NAND | SAB(a1, a0, a2));
}

static const TCGOutOpBinary outop_nand = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_nand,
};

static void tgen_nor(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, NOR | SAB(a1, a0, a2));
}

static const TCGOutOpBinary outop_nor = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_nor,
};

static void tgen_or(TCGContext *s, TCGType type,
                    TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, OR | SAB(a1, a0, a2));
}

static void tgen_ori(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    tcg_out_ori32(s, a0, a1, a2);
}

static const TCGOutOpBinary outop_or = {
    .base.static_constraint = C_O1_I2(r, r, rU),
    .out_rrr = tgen_or,
    .out_rri = tgen_ori,
};

static void tgen_orc(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, ORC | SAB(a1, a0, a2));
}

static const TCGOutOpBinary outop_orc = {
    .base.static_constraint = C_O1_I2(r, r, r),
    .out_rrr = tgen_orc,
};

static TCGConstraintSetIndex cset_mod(TCGType type, unsigned flags)
{
    return have_isa_3_00 ? C_O1_I2(r, r, r) : C_NotImplemented;
}

static void tgen_rems(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? MODSW : MODSD;
    tcg_out32(s, insn | TAB(a0, a1, a2));
}

static const TCGOutOpBinary outop_rems = {
    .base.static_constraint = C_Dynamic,
    .base.dynamic_constraint = cset_mod,
    .out_rrr = tgen_rems,
};

static void tgen_remu(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? MODUW : MODUD;
    tcg_out32(s, insn | TAB(a0, a1, a2));
}

static const TCGOutOpBinary outop_remu = {
    .base.static_constraint = C_Dynamic,
    .base.dynamic_constraint = cset_mod,
    .out_rrr = tgen_remu,
};

static void tgen_rotl(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    if (type == TCG_TYPE_I32) {
        tcg_out32(s, RLWNM | SAB(a1, a0, a2) | MB(0) | ME(31));
    } else {
        tcg_out32(s, RLDCL | SAB(a1, a0, a2) | MB64(0));
    }
}

static void tgen_rotli(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    if (type == TCG_TYPE_I32) {
        tcg_out_rlw(s, RLWINM, a0, a1, a2, 0, 31);
    } else {
        tcg_out_rld(s, RLDICL, a0, a1, a2, 0);
    }
}

static const TCGOutOpBinary outop_rotl = {
    .base.static_constraint = C_O1_I2(r, r, ri),
    .out_rrr = tgen_rotl,
    .out_rri = tgen_rotli,
};

static const TCGOutOpBinary outop_rotr = {
    .base.static_constraint = C_NotImplemented,
};

static void tgen_sar(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? SRAW : SRAD;
    tcg_out32(s, insn | SAB(a1, a0, a2));
}

static void tgen_sari(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    /* Limit immediate shift count lest we create an illegal insn.  */
    if (type == TCG_TYPE_I32) {
        tcg_out_sari32(s, a0, a1, a2 & 31);
    } else {
        tcg_out_sari64(s, a0, a1, a2 & 63);
    }
}

static const TCGOutOpBinary outop_sar = {
    .base.static_constraint = C_O1_I2(r, r, ri),
    .out_rrr = tgen_sar,
    .out_rri = tgen_sari,
};

static void tgen_shl(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? SLW : SLD;
    tcg_out32(s, insn | SAB(a1, a0, a2));
}

static void tgen_shli(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    /* Limit immediate shift count lest we create an illegal insn.  */
    if (type == TCG_TYPE_I32) {
        tcg_out_shli32(s, a0, a1, a2 & 31);
    } else {
        tcg_out_shli64(s, a0, a1, a2 & 63);
    }
}

static const TCGOutOpBinary outop_shl = {
    .base.static_constraint = C_O1_I2(r, r, ri),
    .out_rrr = tgen_shl,
    .out_rri = tgen_shli,
};

static void tgen_shr(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    uint32_t insn = type == TCG_TYPE_I32 ? SRW : SRD;
    tcg_out32(s, insn | SAB(a1, a0, a2));
}

static void tgen_shri(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    /* Limit immediate shift count lest we create an illegal insn.  */
    if (type == TCG_TYPE_I32) {
        tcg_out_shri32(s, a0, a1, a2 & 31);
    } else {
        tcg_out_shri64(s, a0, a1, a2 & 63);
    }
}

static const TCGOutOpBinary outop_shr = {
    .base.static_constraint = C_O1_I2(r, r, ri),
    .out_rrr = tgen_shr,
    .out_rri = tgen_shri,
};

static void tgen_sub(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, SUBF | TAB(a0, a2, a1));
}

static void tgen_subfi(TCGContext *s, TCGType type,
                       TCGReg a0, tcg_target_long a1, TCGReg a2)
{
    tcg_out32(s, SUBFIC | TAI(a0, a2, a1));
}

static const TCGOutOpSubtract outop_sub = {
    .base.static_constraint = C_O1_I2(r, rI, r),
    .out_rrr = tgen_sub,
    .out_rir = tgen_subfi,
};

static void tgen_xor(TCGContext *s, TCGType type,
                     TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, XOR | SAB(a1, a0, a2));
}

static void tgen_xori(TCGContext *s, TCGType type,
                      TCGReg a0, TCGReg a1, tcg_target_long a2)
{
    tcg_out_xori32(s, a0, a1, a2);
}

static const TCGOutOpBinary outop_xor = {
    .base.static_constraint = C_O1_I2(r, r, rU),
    .out_rrr = tgen_xor,
    .out_rri = tgen_xori,
};

static void tgen_bswap16(TCGContext *s, TCGType type,
                         TCGReg dst, TCGReg src, unsigned flags)
{
    TCGReg tmp = dst == src ? TCG_REG_R0 : dst;

    if (have_isa_3_10) {
        tcg_out32(s, BRH | RA(dst) | RS(src));
        if (flags & TCG_BSWAP_OS) {
            tcg_out_ext16s(s, TCG_TYPE_REG, dst, dst);
        } else if ((flags & (TCG_BSWAP_IZ | TCG_BSWAP_OZ)) == TCG_BSWAP_OZ) {
            tcg_out_ext16u(s, dst, dst);
        }
        return;
    }

    /*
     * In the following,
     *   dep(a, b, m) -> (a & ~m) | (b & m)
     *
     * Begin with:                              src = xxxxabcd
     */
    /* tmp = rol32(src, 24) & 0x000000ff            = 0000000c */
    tcg_out_rlw(s, RLWINM, tmp, src, 24, 24, 31);
    /* tmp = dep(tmp, rol32(src, 8), 0x0000ff00)    = 000000dc */
    tcg_out_rlw(s, RLWIMI, tmp, src, 8, 16, 23);

    if (flags & TCG_BSWAP_OS) {
        tcg_out_ext16s(s, TCG_TYPE_REG, dst, tmp);
    } else {
        tcg_out_mov(s, TCG_TYPE_REG, dst, tmp);
    }
}

static const TCGOutOpBswap outop_bswap16 = {
    .base.static_constraint = C_O1_I1(r, r),
    .out_rr = tgen_bswap16,
};

static void tgen_neg(TCGContext *s, TCGType type, TCGReg a0, TCGReg a1)
{
    tcg_out32(s, NEG | RT(a0) | RA(a1));
}

static const TCGOutOpUnary outop_neg = {
    .base.static_constraint = C_O1_I1(r, r),
    .out_rr = tgen_neg,
};

static void tgen_not(TCGContext *s, TCGType type, TCGReg a0, TCGReg a1)
{
    tgen_nor(s, type, a0, a1, a1);
}

static const TCGOutOpUnary outop_not = {
    .base.static_constraint = C_O1_I1(r, r),
    .out_rr = tgen_not,
};


static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type,
                       const TCGArg args[TCG_MAX_OP_ARGS],
                       const int const_args[TCG_MAX_OP_ARGS])
{
    TCGArg a0, a1;

    switch (opc) {
    case INDEX_op_goto_ptr:
        tcg_out32(s, MTSPR | RS(args[0]) | CTR);
        tcg_out32(s, ADDI | TAI(TCG_REG_R3, 0, 0));
        tcg_out32(s, BCCTR | BO_ALWAYS);
        break;
    case INDEX_op_br:
        {
            TCGLabel *l = arg_label(args[0]);
            uint32_t insn = B;

            if (l->has_value) {
                insn |= reloc_pc24_val(tcg_splitwx_to_rx(s->code_ptr),
                                       l->u.value_ptr);
            } else {
                tcg_out_reloc(s, s->code_ptr, R_PPC_REL24, l, 0);
            }
            tcg_out32(s, insn);
        }
        break;
    case INDEX_op_ld8u_i32:
    case INDEX_op_ld8u_i64:
        tcg_out_mem_long(s, LBZ, LBZX, args[0], args[1], args[2]);
        break;
    case INDEX_op_ld8s_i32:
    case INDEX_op_ld8s_i64:
        tcg_out_mem_long(s, LBZ, LBZX, args[0], args[1], args[2]);
        tcg_out_ext8s(s, TCG_TYPE_REG, args[0], args[0]);
        break;
    case INDEX_op_ld16u_i32:
    case INDEX_op_ld16u_i64:
        tcg_out_mem_long(s, LHZ, LHZX, args[0], args[1], args[2]);
        break;
    case INDEX_op_ld16s_i32:
    case INDEX_op_ld16s_i64:
        tcg_out_mem_long(s, LHA, LHAX, args[0], args[1], args[2]);
        break;
    case INDEX_op_ld_i32:
    case INDEX_op_ld32u_i64:
        tcg_out_mem_long(s, LWZ, LWZX, args[0], args[1], args[2]);
        break;
    case INDEX_op_ld32s_i64:
        tcg_out_mem_long(s, LWA, LWAX, args[0], args[1], args[2]);
        break;
    case INDEX_op_ld_i64:
        tcg_out_mem_long(s, LD, LDX, args[0], args[1], args[2]);
        break;
    case INDEX_op_st8_i32:
    case INDEX_op_st8_i64:
        tcg_out_mem_long(s, STB, STBX, args[0], args[1], args[2]);
        break;
    case INDEX_op_st16_i32:
    case INDEX_op_st16_i64:
        tcg_out_mem_long(s, STH, STHX, args[0], args[1], args[2]);
        break;
    case INDEX_op_st_i32:
    case INDEX_op_st32_i64:
        tcg_out_mem_long(s, STW, STWX, args[0], args[1], args[2]);
        break;
    case INDEX_op_st_i64:
        tcg_out_mem_long(s, STD, STDX, args[0], args[1], args[2]);
        break;

    case INDEX_op_qemu_ld_i32:
        tcg_out_qemu_ld(s, args[0], -1, args[1], args[2], TCG_TYPE_I32);
        break;
    case INDEX_op_qemu_ld_i64:
        if (TCG_TARGET_REG_BITS == 64) {
            tcg_out_qemu_ld(s, args[0], -1, args[1], args[2], TCG_TYPE_I64);
        } else {
            tcg_out_qemu_ld(s, args[0], args[1], args[2],
                            args[3], TCG_TYPE_I64);
        }
        break;
    case INDEX_op_qemu_ld_i128:
        tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
        tcg_out_qemu_ldst_i128(s, args[0], args[1], args[2], args[3], true);
        break;

    case INDEX_op_qemu_st_i32:
        tcg_out_qemu_st(s, args[0], -1, args[1], args[2], TCG_TYPE_I32);
        break;
    case INDEX_op_qemu_st_i64:
        if (TCG_TARGET_REG_BITS == 64) {
            tcg_out_qemu_st(s, args[0], -1, args[1], args[2], TCG_TYPE_I64);
        } else {
            tcg_out_qemu_st(s, args[0], args[1], args[2],
                            args[3], TCG_TYPE_I64);
        }
        break;
    case INDEX_op_qemu_st_i128:
        tcg_debug_assert(TCG_TARGET_REG_BITS == 64);
        tcg_out_qemu_ldst_i128(s, args[0], args[1], args[2], args[3], false);
        break;

    case INDEX_op_bswap32_i32:
        tcg_out_bswap32(s, args[0], args[1], 0);
        break;
    case INDEX_op_bswap32_i64:
        tcg_out_bswap32(s, args[0], args[1], args[2]);
        break;
    case INDEX_op_bswap64_i64:
        tcg_out_bswap64(s, args[0], args[1]);
        break;

    case INDEX_op_deposit_i32:
        if (const_args[2]) {
            uint32_t mask = ((2u << (args[4] - 1)) - 1) << args[3];
            tcg_out_andi32(s, args[0], args[0], ~mask);
        } else {
            tcg_out_rlw(s, RLWIMI, args[0], args[2], args[3],
                        32 - args[3] - args[4], 31 - args[3]);
        }
        break;
    case INDEX_op_deposit_i64:
        if (const_args[2]) {
            uint64_t mask = ((2ull << (args[4] - 1)) - 1) << args[3];
            tcg_out_andi64(s, args[0], args[0], ~mask);
        } else {
            tcg_out_rld(s, RLDIMI, args[0], args[2], args[3],
                        64 - args[3] - args[4]);
        }
        break;

    case INDEX_op_extract_i32:
        if (args[2] == 0 && args[3] <= 16) {
            tcg_out32(s, ANDI | SAI(args[1], args[0], (1 << args[3]) - 1));
            break;
        }
        tcg_out_rlw(s, RLWINM, args[0], args[1],
                    32 - args[2], 32 - args[3], 31);
        break;
    case INDEX_op_extract_i64:
        if (args[2] == 0 && args[3] <= 16) {
            tcg_out32(s, ANDI | SAI(args[1], args[0], (1 << args[3]) - 1));
            break;
        }
        tcg_out_rld(s, RLDICL, args[0], args[1], 64 - args[2], 64 - args[3]);
        break;

    case INDEX_op_sextract_i64:
        if (args[2] + args[3] == 32) {
            if (args[2] == 0) {
                tcg_out_ext32s(s, args[0], args[1]);
            } else {
                tcg_out_sari32(s, args[0], args[1], args[2]);
            }
            break;
        }
        /* FALLTHRU */
    case INDEX_op_sextract_i32:
        if (args[2] == 0 && args[3] == 8) {
            tcg_out_ext8s(s, TCG_TYPE_I32, args[0], args[1]);
        } else if (args[2] == 0 && args[3] == 16) {
            tcg_out_ext16s(s, TCG_TYPE_I32, args[0], args[1]);
        } else {
            g_assert_not_reached();
        }
        break;

#if TCG_TARGET_REG_BITS == 64
    case INDEX_op_add2_i64:
#else
    case INDEX_op_add2_i32:
#endif
        /* Note that the CA bit is defined based on the word size of the
           environment.  So in 64-bit mode it's always carry-out of bit 63.
           The fallback code using deposit works just as well for 32-bit.  */
        a0 = args[0], a1 = args[1];
        if (a0 == args[3] || (!const_args[5] && a0 == args[5])) {
            a0 = TCG_REG_R0;
        }
        if (const_args[4]) {
            tcg_out32(s, ADDIC | TAI(a0, args[2], args[4]));
        } else {
            tcg_out32(s, ADDC | TAB(a0, args[2], args[4]));
        }
        if (const_args[5]) {
            tcg_out32(s, (args[5] ? ADDME : ADDZE) | RT(a1) | RA(args[3]));
        } else {
            tcg_out32(s, ADDE | TAB(a1, args[3], args[5]));
        }
        if (a0 != args[0]) {
            tcg_out_mov(s, TCG_TYPE_REG, args[0], a0);
        }
        break;

#if TCG_TARGET_REG_BITS == 64
    case INDEX_op_sub2_i64:
#else
    case INDEX_op_sub2_i32:
#endif
        a0 = args[0], a1 = args[1];
        if (a0 == args[5] || (!const_args[3] && a0 == args[3])) {
            a0 = TCG_REG_R0;
        }
        if (const_args[2]) {
            tcg_out32(s, SUBFIC | TAI(a0, args[4], args[2]));
        } else {
            tcg_out32(s, SUBFC | TAB(a0, args[4], args[2]));
        }
        if (const_args[3]) {
            tcg_out32(s, (args[3] ? SUBFME : SUBFZE) | RT(a1) | RA(args[5]));
        } else {
            tcg_out32(s, SUBFE | TAB(a1, args[5], args[3]));
        }
        if (a0 != args[0]) {
            tcg_out_mov(s, TCG_TYPE_REG, args[0], a0);
        }
        break;

    case INDEX_op_mb:
        tcg_out_mb(s, args[0]);
        break;

    case INDEX_op_call:      /* Always emitted via tcg_out_call.  */
    case INDEX_op_exit_tb:   /* Always emitted via tcg_out_exit_tb.  */
    case INDEX_op_goto_tb:   /* Always emitted via tcg_out_goto_tb.  */
    case INDEX_op_ext_i32_i64:  /* Always emitted via tcg_reg_alloc_op.  */
    case INDEX_op_extu_i32_i64:
    case INDEX_op_extrl_i64_i32:
    default:
        g_assert_not_reached();
    }
}

int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece)
{
    switch (opc) {
    case INDEX_op_and_vec:
    case INDEX_op_or_vec:
    case INDEX_op_xor_vec:
    case INDEX_op_andc_vec:
    case INDEX_op_not_vec:
    case INDEX_op_nor_vec:
    case INDEX_op_eqv_vec:
    case INDEX_op_nand_vec:
        return 1;
    case INDEX_op_orc_vec:
        return have_isa_2_07;
    case INDEX_op_add_vec:
    case INDEX_op_sub_vec:
    case INDEX_op_smax_vec:
    case INDEX_op_smin_vec:
    case INDEX_op_umax_vec:
    case INDEX_op_umin_vec:
    case INDEX_op_shlv_vec:
    case INDEX_op_shrv_vec:
    case INDEX_op_sarv_vec:
    case INDEX_op_rotlv_vec:
        return vece <= MO_32 || have_isa_2_07;
    case INDEX_op_ssadd_vec:
    case INDEX_op_sssub_vec:
    case INDEX_op_usadd_vec:
    case INDEX_op_ussub_vec:
        return vece <= MO_32;
    case INDEX_op_shli_vec:
    case INDEX_op_shri_vec:
    case INDEX_op_sari_vec:
    case INDEX_op_rotli_vec:
        return vece <= MO_32 || have_isa_2_07 ? -1 : 0;
    case INDEX_op_cmp_vec:
    case INDEX_op_cmpsel_vec:
        return vece <= MO_32 || have_isa_2_07 ? 1 : 0;
    case INDEX_op_neg_vec:
        return vece >= MO_32 && have_isa_3_00;
    case INDEX_op_mul_vec:
        switch (vece) {
        case MO_8:
        case MO_16:
            return -1;
        case MO_32:
            return have_isa_2_07 ? 1 : -1;
        case MO_64:
            return have_isa_3_10;
        }
        return 0;
    case INDEX_op_bitsel_vec:
        return have_vsx;
    case INDEX_op_rotrv_vec:
        return -1;
    default:
        return 0;
    }
}

static bool tcg_out_dup_vec(TCGContext *s, TCGType type, unsigned vece,
                            TCGReg dst, TCGReg src)
{
    tcg_debug_assert(dst >= TCG_REG_V0);

    /* Splat from integer reg allowed via constraints for v3.00.  */
    if (src < TCG_REG_V0) {
        tcg_debug_assert(have_isa_3_00);
        switch (vece) {
        case MO_64:
            tcg_out32(s, MTVSRDD | VRT(dst) | RA(src) | RB(src));
            return true;
        case MO_32:
            tcg_out32(s, MTVSRWS | VRT(dst) | RA(src));
            return true;
        default:
            /* Fail, so that we fall back on either dupm or mov+dup.  */
            return false;
        }
    }

    /*
     * Recall we use (or emulate) VSX integer loads, so the integer is
     * right justified within the left (zero-index) double-word.
     */
    switch (vece) {
    case MO_8:
        tcg_out32(s, VSPLTB | VRT(dst) | VRB(src) | (7 << 16));
        break;
    case MO_16:
        tcg_out32(s, VSPLTH | VRT(dst) | VRB(src) | (3 << 16));
        break;
    case MO_32:
        tcg_out32(s, VSPLTW | VRT(dst) | VRB(src) | (1 << 16));
        break;
    case MO_64:
        if (have_vsx) {
            tcg_out32(s, XXPERMDI | VRT(dst) | VRA(src) | VRB(src));
            break;
        }
        tcg_out_vsldoi(s, TCG_VEC_TMP1, src, src, 8);
        tcg_out_vsldoi(s, dst, TCG_VEC_TMP1, src, 8);
        break;
    default:
        g_assert_not_reached();
    }
    return true;
}

static bool tcg_out_dupm_vec(TCGContext *s, TCGType type, unsigned vece,
                             TCGReg out, TCGReg base, intptr_t offset)
{
    int elt;

    tcg_debug_assert(out >= TCG_REG_V0);
    switch (vece) {
    case MO_8:
        if (have_isa_3_00) {
            tcg_out_mem_long(s, LXV, LVX, out, base, offset & -16);
        } else {
            tcg_out_mem_long(s, 0, LVEBX, out, base, offset);
        }
        elt = extract32(offset, 0, 4);
#if !HOST_BIG_ENDIAN
        elt ^= 15;
#endif
        tcg_out32(s, VSPLTB | VRT(out) | VRB(out) | (elt << 16));
        break;
    case MO_16:
        tcg_debug_assert((offset & 1) == 0);
        if (have_isa_3_00) {
            tcg_out_mem_long(s, LXV | 8, LVX, out, base, offset & -16);
        } else {
            tcg_out_mem_long(s, 0, LVEHX, out, base, offset);
        }
        elt = extract32(offset, 1, 3);
#if !HOST_BIG_ENDIAN
        elt ^= 7;
#endif
        tcg_out32(s, VSPLTH | VRT(out) | VRB(out) | (elt << 16));
        break;
    case MO_32:
        if (have_isa_3_00) {
            tcg_out_mem_long(s, 0, LXVWSX, out, base, offset);
            break;
        }
        tcg_debug_assert((offset & 3) == 0);
        tcg_out_mem_long(s, 0, LVEWX, out, base, offset);
        elt = extract32(offset, 2, 2);
#if !HOST_BIG_ENDIAN
        elt ^= 3;
#endif
        tcg_out32(s, VSPLTW | VRT(out) | VRB(out) | (elt << 16));
        break;
    case MO_64:
        if (have_vsx) {
            tcg_out_mem_long(s, 0, LXVDSX, out, base, offset);
            break;
        }
        tcg_debug_assert((offset & 7) == 0);
        tcg_out_mem_long(s, 0, LVX, out, base, offset & -16);
        tcg_out_vsldoi(s, TCG_VEC_TMP1, out, out, 8);
        elt = extract32(offset, 3, 1);
#if !HOST_BIG_ENDIAN
        elt = !elt;
#endif
        if (elt) {
            tcg_out_vsldoi(s, out, out, TCG_VEC_TMP1, 8);
        } else {
            tcg_out_vsldoi(s, out, TCG_VEC_TMP1, out, 8);
        }
        break;
    default:
        g_assert_not_reached();
    }
    return true;
}

static void tcg_out_not_vec(TCGContext *s, TCGReg a0, TCGReg a1)
{
    tcg_out32(s, VNOR | VRT(a0) | VRA(a1) | VRB(a1));
}

static void tcg_out_or_vec(TCGContext *s, TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, VOR | VRT(a0) | VRA(a1) | VRB(a2));
}

static void tcg_out_orc_vec(TCGContext *s, TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, VORC | VRT(a0) | VRA(a1) | VRB(a2));
}

static void tcg_out_and_vec(TCGContext *s, TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, VAND | VRT(a0) | VRA(a1) | VRB(a2));
}

static void tcg_out_andc_vec(TCGContext *s, TCGReg a0, TCGReg a1, TCGReg a2)
{
    tcg_out32(s, VANDC | VRT(a0) | VRA(a1) | VRB(a2));
}

static void tcg_out_bitsel_vec(TCGContext *s, TCGReg d,
                               TCGReg c, TCGReg t, TCGReg f)
{
    if (TCG_TARGET_HAS_bitsel_vec) {
        tcg_out32(s, XXSEL | VRT(d) | VRC(c) | VRB(t) | VRA(f));
    } else {
        tcg_out_and_vec(s, TCG_VEC_TMP2, t, c);
        tcg_out_andc_vec(s, d, f, c);
        tcg_out_or_vec(s, d, d, TCG_VEC_TMP2);
    }
}

static bool tcg_out_cmp_vec_noinv(TCGContext *s, unsigned vece, TCGReg a0,
                                  TCGReg a1, TCGReg a2, TCGCond cond)
{
    static const uint32_t
        eq_op[4]  = { VCMPEQUB, VCMPEQUH, VCMPEQUW, VCMPEQUD },
        ne_op[4]  = { VCMPNEB, VCMPNEH, VCMPNEW, 0 },
        gts_op[4] = { VCMPGTSB, VCMPGTSH, VCMPGTSW, VCMPGTSD },
        gtu_op[4] = { VCMPGTUB, VCMPGTUH, VCMPGTUW, VCMPGTUD };
    uint32_t insn;

    bool need_swap = false, need_inv = false;

    tcg_debug_assert(vece <= MO_32 || have_isa_2_07);

    switch (cond) {
    case TCG_COND_EQ:
    case TCG_COND_GT:
    case TCG_COND_GTU:
        break;
    case TCG_COND_NE:
        if (have_isa_3_00 && vece <= MO_32) {
            break;
        }
        /* fall through */
    case TCG_COND_LE:
    case TCG_COND_LEU:
        need_inv = true;
        break;
    case TCG_COND_LT:
    case TCG_COND_LTU:
        need_swap = true;
        break;
    case TCG_COND_GE:
    case TCG_COND_GEU:
        need_swap = need_inv = true;
        break;
    default:
        g_assert_not_reached();
    }

    if (need_inv) {
        cond = tcg_invert_cond(cond);
    }
    if (need_swap) {
        TCGReg swap = a1;
        a1 = a2;
        a2 = swap;
        cond = tcg_swap_cond(cond);
    }

    switch (cond) {
    case TCG_COND_EQ:
        insn = eq_op[vece];
        break;
    case TCG_COND_NE:
        insn = ne_op[vece];
        break;
    case TCG_COND_GT:
        insn = gts_op[vece];
        break;
    case TCG_COND_GTU:
        insn = gtu_op[vece];
        break;
    default:
        g_assert_not_reached();
    }
    tcg_out32(s, insn | VRT(a0) | VRA(a1) | VRB(a2));

    return need_inv;
}

static void tcg_out_cmp_vec(TCGContext *s, unsigned vece, TCGReg a0,
                            TCGReg a1, TCGReg a2, TCGCond cond)
{
    if (tcg_out_cmp_vec_noinv(s, vece, a0, a1, a2, cond)) {
        tcg_out_not_vec(s, a0, a0);
    }
}

static void tcg_out_cmpsel_vec(TCGContext *s, unsigned vece, TCGReg a0,
                               TCGReg c1, TCGReg c2, TCGArg v3, int const_v3,
                               TCGReg v4, TCGCond cond)
{
    bool inv = tcg_out_cmp_vec_noinv(s, vece, TCG_VEC_TMP1, c1, c2, cond);

    if (!const_v3) {
        if (inv) {
            tcg_out_bitsel_vec(s, a0, TCG_VEC_TMP1, v4, v3);
        } else {
            tcg_out_bitsel_vec(s, a0, TCG_VEC_TMP1, v3, v4);
        }
    } else if (v3) {
        if (inv) {
            tcg_out_orc_vec(s, a0, v4, TCG_VEC_TMP1);
        } else {
            tcg_out_or_vec(s, a0, v4, TCG_VEC_TMP1);
        }
    } else {
        if (inv) {
            tcg_out_and_vec(s, a0, v4, TCG_VEC_TMP1);
        } else {
            tcg_out_andc_vec(s, a0, v4, TCG_VEC_TMP1);
        }
    }
}

static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc,
                           unsigned vecl, unsigned vece,
                           const TCGArg args[TCG_MAX_OP_ARGS],
                           const int const_args[TCG_MAX_OP_ARGS])
{
    static const uint32_t
        add_op[4] = { VADDUBM, VADDUHM, VADDUWM, VADDUDM },
        sub_op[4] = { VSUBUBM, VSUBUHM, VSUBUWM, VSUBUDM },
        mul_op[4] = { 0, 0, VMULUWM, VMULLD },
        neg_op[4] = { 0, 0, VNEGW, VNEGD },
        ssadd_op[4] = { VADDSBS, VADDSHS, VADDSWS, 0 },
        usadd_op[4] = { VADDUBS, VADDUHS, VADDUWS, 0 },
        sssub_op[4] = { VSUBSBS, VSUBSHS, VSUBSWS, 0 },
        ussub_op[4] = { VSUBUBS, VSUBUHS, VSUBUWS, 0 },
        umin_op[4] = { VMINUB, VMINUH, VMINUW, VMINUD },
        smin_op[4] = { VMINSB, VMINSH, VMINSW, VMINSD },
        umax_op[4] = { VMAXUB, VMAXUH, VMAXUW, VMAXUD },
        smax_op[4] = { VMAXSB, VMAXSH, VMAXSW, VMAXSD },
        shlv_op[4] = { VSLB, VSLH, VSLW, VSLD },
        shrv_op[4] = { VSRB, VSRH, VSRW, VSRD },
        sarv_op[4] = { VSRAB, VSRAH, VSRAW, VSRAD },
        mrgh_op[4] = { VMRGHB, VMRGHH, VMRGHW, 0 },
        mrgl_op[4] = { VMRGLB, VMRGLH, VMRGLW, 0 },
        muleu_op[4] = { VMULEUB, VMULEUH, VMULEUW, 0 },
        mulou_op[4] = { VMULOUB, VMULOUH, VMULOUW, 0 },
        pkum_op[4] = { VPKUHUM, VPKUWUM, 0, 0 },
        rotl_op[4] = { VRLB, VRLH, VRLW, VRLD };

    TCGType type = vecl + TCG_TYPE_V64;
    TCGArg a0 = args[0], a1 = args[1], a2 = args[2];
    uint32_t insn;

    switch (opc) {
    case INDEX_op_ld_vec:
        tcg_out_ld(s, type, a0, a1, a2);
        return;
    case INDEX_op_st_vec:
        tcg_out_st(s, type, a0, a1, a2);
        return;
    case INDEX_op_dupm_vec:
        tcg_out_dupm_vec(s, type, vece, a0, a1, a2);
        return;

    case INDEX_op_add_vec:
        insn = add_op[vece];
        break;
    case INDEX_op_sub_vec:
        insn = sub_op[vece];
        break;
    case INDEX_op_neg_vec:
        insn = neg_op[vece];
        a2 = a1;
        a1 = 0;
        break;
    case INDEX_op_mul_vec:
        insn = mul_op[vece];
        break;
    case INDEX_op_ssadd_vec:
        insn = ssadd_op[vece];
        break;
    case INDEX_op_sssub_vec:
        insn = sssub_op[vece];
        break;
    case INDEX_op_usadd_vec:
        insn = usadd_op[vece];
        break;
    case INDEX_op_ussub_vec:
        insn = ussub_op[vece];
        break;
    case INDEX_op_smin_vec:
        insn = smin_op[vece];
        break;
    case INDEX_op_umin_vec:
        insn = umin_op[vece];
        break;
    case INDEX_op_smax_vec:
        insn = smax_op[vece];
        break;
    case INDEX_op_umax_vec:
        insn = umax_op[vece];
        break;
    case INDEX_op_shlv_vec:
        insn = shlv_op[vece];
        break;
    case INDEX_op_shrv_vec:
        insn = shrv_op[vece];
        break;
    case INDEX_op_sarv_vec:
        insn = sarv_op[vece];
        break;
    case INDEX_op_and_vec:
        tcg_out_and_vec(s, a0, a1, a2);
        return;
    case INDEX_op_or_vec:
        tcg_out_or_vec(s, a0, a1, a2);
        return;
    case INDEX_op_xor_vec:
        insn = VXOR;
        break;
    case INDEX_op_andc_vec:
        tcg_out_andc_vec(s, a0, a1, a2);
        return;
    case INDEX_op_not_vec:
        tcg_out_not_vec(s, a0, a1);
        return;
    case INDEX_op_orc_vec:
        tcg_out_orc_vec(s, a0, a1, a2);
        return;
    case INDEX_op_nand_vec:
        insn = VNAND;
        break;
    case INDEX_op_nor_vec:
        insn = VNOR;
        break;
    case INDEX_op_eqv_vec:
        insn = VEQV;
        break;

    case INDEX_op_cmp_vec:
        tcg_out_cmp_vec(s, vece, a0, a1, a2, args[3]);
        return;
    case INDEX_op_cmpsel_vec:
        tcg_out_cmpsel_vec(s, vece, a0, a1, a2,
                           args[3], const_args[3], args[4], args[5]);
        return;
    case INDEX_op_bitsel_vec:
        tcg_out_bitsel_vec(s, a0, a1, a2, args[3]);
        return;

    case INDEX_op_dup2_vec:
        assert(TCG_TARGET_REG_BITS == 32);
        /* With inputs a1 = xLxx, a2 = xHxx  */
        tcg_out32(s, VMRGHW | VRT(a0) | VRA(a2) | VRB(a1));  /* a0  = xxHL */
        tcg_out_vsldoi(s, TCG_VEC_TMP1, a0, a0, 8);          /* tmp = HLxx */
        tcg_out_vsldoi(s, a0, a0, TCG_VEC_TMP1, 8);          /* a0  = HLHL */
        return;

    case INDEX_op_ppc_mrgh_vec:
        insn = mrgh_op[vece];
        break;
    case INDEX_op_ppc_mrgl_vec:
        insn = mrgl_op[vece];
        break;
    case INDEX_op_ppc_muleu_vec:
        insn = muleu_op[vece];
        break;
    case INDEX_op_ppc_mulou_vec:
        insn = mulou_op[vece];
        break;
    case INDEX_op_ppc_pkum_vec:
        insn = pkum_op[vece];
        break;
    case INDEX_op_rotlv_vec:
        insn = rotl_op[vece];
        break;
    case INDEX_op_ppc_msum_vec:
        tcg_debug_assert(vece == MO_16);
        tcg_out32(s, VMSUMUHM | VRT(a0) | VRA(a1) | VRB(a2) | VRC(args[3]));
        return;

    case INDEX_op_mov_vec:  /* Always emitted via tcg_out_mov.  */
    case INDEX_op_dup_vec:  /* Always emitted via tcg_out_dup_vec.  */
    default:
        g_assert_not_reached();
    }

    tcg_debug_assert(insn != 0);
    tcg_out32(s, insn | VRT(a0) | VRA(a1) | VRB(a2));
}

static void expand_vec_shi(TCGType type, unsigned vece, TCGv_vec v0,
                           TCGv_vec v1, TCGArg imm, TCGOpcode opci)
{
    TCGv_vec t1;

    if (vece == MO_32) {
        /*
         * Only 5 bits are significant, and VSPLTISB can represent -16..15.
         * So using negative numbers gets us the 4th bit easily.
         */
        imm = sextract32(imm, 0, 5);
    } else {
        imm &= (8 << vece) - 1;
    }

    /* Splat w/bytes for xxspltib when 2.07 allows MO_64. */
    t1 = tcg_constant_vec(type, MO_8, imm);
    vec_gen_3(opci, type, vece, tcgv_vec_arg(v0),
              tcgv_vec_arg(v1), tcgv_vec_arg(t1));
}

static void expand_vec_mul(TCGType type, unsigned vece, TCGv_vec v0,
                           TCGv_vec v1, TCGv_vec v2)
{
    TCGv_vec t1 = tcg_temp_new_vec(type);
    TCGv_vec t2 = tcg_temp_new_vec(type);
    TCGv_vec c0, c16;

    switch (vece) {
    case MO_8:
    case MO_16:
        vec_gen_3(INDEX_op_ppc_muleu_vec, type, vece, tcgv_vec_arg(t1),
                  tcgv_vec_arg(v1), tcgv_vec_arg(v2));
        vec_gen_3(INDEX_op_ppc_mulou_vec, type, vece, tcgv_vec_arg(t2),
                  tcgv_vec_arg(v1), tcgv_vec_arg(v2));
        vec_gen_3(INDEX_op_ppc_mrgh_vec, type, vece + 1, tcgv_vec_arg(v0),
                  tcgv_vec_arg(t1), tcgv_vec_arg(t2));
        vec_gen_3(INDEX_op_ppc_mrgl_vec, type, vece + 1, tcgv_vec_arg(t1),
                  tcgv_vec_arg(t1), tcgv_vec_arg(t2));
        vec_gen_3(INDEX_op_ppc_pkum_vec, type, vece, tcgv_vec_arg(v0),
                  tcgv_vec_arg(v0), tcgv_vec_arg(t1));
        break;

    case MO_32:
        tcg_debug_assert(!have_isa_2_07);
        /*
         * Only 5 bits are significant, and VSPLTISB can represent -16..15.
         * So using -16 is a quick way to represent 16.
         */
        c16 = tcg_constant_vec(type, MO_8, -16);
        c0 = tcg_constant_vec(type, MO_8, 0);

        vec_gen_3(INDEX_op_rotlv_vec, type, MO_32, tcgv_vec_arg(t1),
                  tcgv_vec_arg(v2), tcgv_vec_arg(c16));
        vec_gen_3(INDEX_op_ppc_mulou_vec, type, MO_16, tcgv_vec_arg(t2),
                  tcgv_vec_arg(v1), tcgv_vec_arg(v2));
        vec_gen_4(INDEX_op_ppc_msum_vec, type, MO_16, tcgv_vec_arg(t1),
                  tcgv_vec_arg(v1), tcgv_vec_arg(t1), tcgv_vec_arg(c0));
        vec_gen_3(INDEX_op_shlv_vec, type, MO_32, tcgv_vec_arg(t1),
                  tcgv_vec_arg(t1), tcgv_vec_arg(c16));
        tcg_gen_add_vec(MO_32, v0, t1, t2);
        break;

    default:
        g_assert_not_reached();
    }
    tcg_temp_free_vec(t1);
    tcg_temp_free_vec(t2);
}

void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece,
                       TCGArg a0, ...)
{
    va_list va;
    TCGv_vec v0, v1, v2, t0;
    TCGArg a2;

    va_start(va, a0);
    v0 = temp_tcgv_vec(arg_temp(a0));
    v1 = temp_tcgv_vec(arg_temp(va_arg(va, TCGArg)));
    a2 = va_arg(va, TCGArg);

    switch (opc) {
    case INDEX_op_shli_vec:
        expand_vec_shi(type, vece, v0, v1, a2, INDEX_op_shlv_vec);
        break;
    case INDEX_op_shri_vec:
        expand_vec_shi(type, vece, v0, v1, a2, INDEX_op_shrv_vec);
        break;
    case INDEX_op_sari_vec:
        expand_vec_shi(type, vece, v0, v1, a2, INDEX_op_sarv_vec);
        break;
    case INDEX_op_rotli_vec:
        expand_vec_shi(type, vece, v0, v1, a2, INDEX_op_rotlv_vec);
        break;
    case INDEX_op_mul_vec:
        v2 = temp_tcgv_vec(arg_temp(a2));
        expand_vec_mul(type, vece, v0, v1, v2);
        break;
    case INDEX_op_rotlv_vec:
        v2 = temp_tcgv_vec(arg_temp(a2));
        t0 = tcg_temp_new_vec(type);
        tcg_gen_neg_vec(vece, t0, v2);
        tcg_gen_rotlv_vec(vece, v0, v1, t0);
        tcg_temp_free_vec(t0);
        break;
    default:
        g_assert_not_reached();
    }
    va_end(va);
}

static TCGConstraintSetIndex
tcg_target_op_def(TCGOpcode op, TCGType type, unsigned flags)
{
    switch (op) {
    case INDEX_op_goto_ptr:
        return C_O0_I1(r);

    case INDEX_op_ld8u_i32:
    case INDEX_op_ld8s_i32:
    case INDEX_op_ld16u_i32:
    case INDEX_op_ld16s_i32:
    case INDEX_op_ld_i32:
    case INDEX_op_bswap32_i32:
    case INDEX_op_extract_i32:
    case INDEX_op_sextract_i32:
    case INDEX_op_ld8u_i64:
    case INDEX_op_ld8s_i64:
    case INDEX_op_ld16u_i64:
    case INDEX_op_ld16s_i64:
    case INDEX_op_ld32u_i64:
    case INDEX_op_ld32s_i64:
    case INDEX_op_ld_i64:
    case INDEX_op_ext_i32_i64:
    case INDEX_op_extu_i32_i64:
    case INDEX_op_bswap32_i64:
    case INDEX_op_bswap64_i64:
    case INDEX_op_extract_i64:
    case INDEX_op_sextract_i64:
        return C_O1_I1(r, r);

    case INDEX_op_st8_i32:
    case INDEX_op_st16_i32:
    case INDEX_op_st_i32:
    case INDEX_op_st8_i64:
    case INDEX_op_st16_i64:
    case INDEX_op_st32_i64:
    case INDEX_op_st_i64:
        return C_O0_I2(r, r);

    case INDEX_op_deposit_i32:
    case INDEX_op_deposit_i64:
        return C_O1_I2(r, 0, rZ);
    case INDEX_op_add2_i64:
    case INDEX_op_add2_i32:
        return C_O2_I4(r, r, r, r, rI, rZM);
    case INDEX_op_sub2_i64:
    case INDEX_op_sub2_i32:
        return C_O2_I4(r, r, rI, rZM, r, r);

    case INDEX_op_qemu_ld_i32:
        return C_O1_I1(r, r);
    case INDEX_op_qemu_ld_i64:
        return TCG_TARGET_REG_BITS == 64 ? C_O1_I1(r, r) : C_O2_I1(r, r, r);

    case INDEX_op_qemu_st_i32:
        return C_O0_I2(r, r);
    case INDEX_op_qemu_st_i64:
        return TCG_TARGET_REG_BITS == 64 ? C_O0_I2(r, r) : C_O0_I3(r, r, r);

    case INDEX_op_qemu_ld_i128:
        return C_N1O1_I1(o, m, r);
    case INDEX_op_qemu_st_i128:
        return C_O0_I3(o, m, r);

    case INDEX_op_add_vec:
    case INDEX_op_sub_vec:
    case INDEX_op_mul_vec:
    case INDEX_op_and_vec:
    case INDEX_op_or_vec:
    case INDEX_op_xor_vec:
    case INDEX_op_andc_vec:
    case INDEX_op_orc_vec:
    case INDEX_op_nor_vec:
    case INDEX_op_eqv_vec:
    case INDEX_op_nand_vec:
    case INDEX_op_cmp_vec:
    case INDEX_op_ssadd_vec:
    case INDEX_op_sssub_vec:
    case INDEX_op_usadd_vec:
    case INDEX_op_ussub_vec:
    case INDEX_op_smax_vec:
    case INDEX_op_smin_vec:
    case INDEX_op_umax_vec:
    case INDEX_op_umin_vec:
    case INDEX_op_shlv_vec:
    case INDEX_op_shrv_vec:
    case INDEX_op_sarv_vec:
    case INDEX_op_rotlv_vec:
    case INDEX_op_rotrv_vec:
    case INDEX_op_ppc_mrgh_vec:
    case INDEX_op_ppc_mrgl_vec:
    case INDEX_op_ppc_muleu_vec:
    case INDEX_op_ppc_mulou_vec:
    case INDEX_op_ppc_pkum_vec:
    case INDEX_op_dup2_vec:
        return C_O1_I2(v, v, v);

    case INDEX_op_not_vec:
    case INDEX_op_neg_vec:
        return C_O1_I1(v, v);

    case INDEX_op_dup_vec:
        return have_isa_3_00 ? C_O1_I1(v, vr) : C_O1_I1(v, v);

    case INDEX_op_ld_vec:
    case INDEX_op_dupm_vec:
        return C_O1_I1(v, r);

    case INDEX_op_st_vec:
        return C_O0_I2(v, r);

    case INDEX_op_bitsel_vec:
    case INDEX_op_ppc_msum_vec:
        return C_O1_I3(v, v, v, v);
    case INDEX_op_cmpsel_vec:
        return C_O1_I4(v, v, v, vZM, v);

    default:
        return C_NotImplemented;
    }
}

static void tcg_target_init(TCGContext *s)
{
    tcg_target_available_regs[TCG_TYPE_I32] = 0xffffffff;
    tcg_target_available_regs[TCG_TYPE_I64] = 0xffffffff;
    if (have_altivec) {
        tcg_target_available_regs[TCG_TYPE_V64] = 0xffffffff00000000ull;
        tcg_target_available_regs[TCG_TYPE_V128] = 0xffffffff00000000ull;
    }

    tcg_target_call_clobber_regs = 0;
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R0);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R2);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R3);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R4);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R5);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R6);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R7);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R8);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R9);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R10);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R11);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R12);

    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V0);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V1);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V2);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V3);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V4);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V5);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V6);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V7);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V8);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V9);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V10);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V11);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V12);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V13);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V14);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V15);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V16);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V17);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V18);
    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V19);

    s->reserved_regs = 0;
    tcg_regset_set_reg(s->reserved_regs, TCG_REG_R0); /* tcg temp */
    tcg_regset_set_reg(s->reserved_regs, TCG_REG_R1); /* stack pointer */
#if defined(_CALL_SYSV)
    tcg_regset_set_reg(s->reserved_regs, TCG_REG_R2); /* toc pointer */
#endif
#if defined(_CALL_SYSV) || TCG_TARGET_REG_BITS == 64
    tcg_regset_set_reg(s->reserved_regs, TCG_REG_R13); /* thread pointer */
#endif
    tcg_regset_set_reg(s->reserved_regs, TCG_REG_TMP1);
    tcg_regset_set_reg(s->reserved_regs, TCG_REG_TMP2);
    tcg_regset_set_reg(s->reserved_regs, TCG_VEC_TMP1);
    tcg_regset_set_reg(s->reserved_regs, TCG_VEC_TMP2);
    if (USE_REG_TB) {
        tcg_regset_set_reg(s->reserved_regs, TCG_REG_TB);  /* tb->tc_ptr */
    }
}

#ifdef __ELF__
typedef struct {
    DebugFrameCIE cie;
    DebugFrameFDEHeader fde;
    uint8_t fde_def_cfa[4];
    uint8_t fde_reg_ofs[ARRAY_SIZE(tcg_target_callee_save_regs) * 2 + 3];
} DebugFrame;

/* We're expecting a 2 byte uleb128 encoded value.  */
QEMU_BUILD_BUG_ON(FRAME_SIZE >= (1 << 14));

#if TCG_TARGET_REG_BITS == 64
# define ELF_HOST_MACHINE EM_PPC64
#else
# define ELF_HOST_MACHINE EM_PPC
#endif

static DebugFrame debug_frame = {
    .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */
    .cie.id = -1,
    .cie.version = 1,
    .cie.code_align = 1,
    .cie.data_align = (-SZR & 0x7f),         /* sleb128 -SZR */
    .cie.return_column = 65,

    /* Total FDE size does not include the "len" member.  */
    .fde.len = sizeof(DebugFrame) - offsetof(DebugFrame, fde.cie_offset),

    .fde_def_cfa = {
        12, TCG_REG_R1,                 /* DW_CFA_def_cfa r1, ... */
        (FRAME_SIZE & 0x7f) | 0x80,     /* ... uleb128 FRAME_SIZE */
        (FRAME_SIZE >> 7)
    },
    .fde_reg_ofs = {
        /* DW_CFA_offset_extended_sf, lr, LR_OFFSET */
        0x11, 65, (LR_OFFSET / -SZR) & 0x7f,
    }
};

void tcg_register_jit(const void *buf, size_t buf_size)
{
    uint8_t *p = &debug_frame.fde_reg_ofs[3];
    int i;

    for (i = 0; i < ARRAY_SIZE(tcg_target_callee_save_regs); ++i, p += 2) {
        p[0] = 0x80 + tcg_target_callee_save_regs[i];
        p[1] = (FRAME_SIZE - (REG_SAVE_BOT + i * SZR)) / SZR;
    }

    debug_frame.fde.func_start = (uintptr_t)buf;
    debug_frame.fde.func_len = buf_size;

    tcg_register_jit_int(buf, buf_size, &debug_frame, sizeof(debug_frame));
}
#endif /* __ELF__ */
#undef VMULEUB
#undef VMULEUH
#undef VMULEUW
#undef VMULOUB
#undef VMULOUH
#undef VMULOUW
#undef VMSUMUHM